CHILD SURVIVAL A Project of The Department of Demography The Australian National University Sponsored by The Ford Foundation
Item
- Title
-
CHILD SURVIVAL
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation - extracted text
-
Research Note on
CHILD
SURVIVAL
Number
IOCS
Date
26 June 1986
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
’i
I
!
I
THE DETERMINANTS OF INFANT AND CHILD MORTALITY
IN KOREA: 1955-1973
•1
* <
Tai-Hun Kim
IPDP Visitor,
Child Survival Project,
International Population Dynamics Program,
Department of Demography,
The Australian National University
■- J
» i
i
i
■j
I
i
«•
' I
Note:
■'" l1*.:?.-. B Y
-J r
) -~
INFORMATION
CEMFE
4
j'
-
Child Survival Research Notes are brief discussions of
issues of current relevance to researchers and policy
makers concerned with problems of high infant and child
mortality in the world. The International Population
Dynamics Program, Department of Demography, The Australian
National University, distributes these notes with their
regular Bibliographic Circular. Production of the Child
Survival Research Notes is made possible through a grant
from the Ford Foundation (840-0893). Responsibility for
the content of Child Survival Research Notes rests
with the author(s) alone, and not the above-listed
organisations .
1
THE DETERMINANTS OF INFANT AND CHILD MORTALITY
IN KOREA: 1955-1973
1• Introduction
Infant and child mortality varies among individual families, partly
because of differences in the demographic characteristics of the
mothers. It has been known for a long time that younger or older women,
higher parity women, and those with less than optimal birth spacing are
more likely to experience infant or child loss (e.g., Wray, 1971;
Bouvier and van der Tak, 1976; Federici and Terrenato,
1980; Arriaga,
1980; Edmonston, Greene and Smith, 1981; Winicoff, 1983; Trussell and
Hammerslough, 1983; Rutstein, 1983; Hobcraft, McDonald and Rutstein,
1983; Ruzicka,
1984; Gubhaju, 1984; de Sweemer, 1984). Furthermore,
previous foetal or infant loss in the family increases the risk that a
pregnant woman will again lose a child (Bouvier and van der Tak, 1976).
In the developing countries previous child loss is considered one of the
most important factors for the survival chances of subsequent children
(Swenson, 1978; Chowdhury, 1981; Gubhaju, 1984).
•phe inverse relationship between socioeconomic status of the parents and
infant mortality reflects differences in parental knowledge and skills
among individual families, as well as variations in family resources
(Frenzen and Hogan, 1982). Father’s education, mother’s education and
father’s occupation each have independent effects upon infant survival
in the developed (MacMahon, Kovar and Feldman, 1972), as well as
developing countries (Caldwell, 1979; Schultz, 1980; Caldwell, Reddy and
Caldwell,
1983). Among those factors, mother's education has been found
in some countries to have greater impact upon infant survival than that
of the father, since mothers normally have more direct responsibilities
for child care (Caldwell, 1979; Cochrane, 1980; Caldwell and McDonald,
1981; Frenzen and Hogan, 1982; Flegg, 1982).
Another important determinant of infant and child mortality is the
mother’s place of residence, which is frequently considered a proxy for
1980; Arriaga and Hobbs, 1982; Hobcraft,
living conditions (Arriaga,
In
many countries urban infant mortality
McDonald and Rutstein, 1983).
has declined at a more rapid pace than in rural areas (Johnson, 1964).
This
study,
therefore,
therefore. identifies the major factors which were^
years) mortality in Korea,
associated with infant and child (age 1-4 years)
examining the demographic and socioeconomic differentials in infant and
child mortality of the five-year birth cohorts during the period 1955-73
based on the 1974 Korean National Fertility Survey (KNFS).
2
2. Methodology
In this analysis we excluded all children who were born less than one
year before the survey for infant mortality and less than five years
before the survey for child mortality, to eliminate the effect of
truncation- We also excluded from the analysis the births that occurred
before 1955 (that is, approximately 15 years before the survey) because
of the uncertainty about completeness of reports on births and deaths
which occurred in the remote past and to reduce the likelihood of source
of the biases (see Park and Park,
1981; Ruzicka, 1984; Kim,
1986).
Therefore, the present study includes children born between one and
approximately 15 years before the survey for infant mortality and
between 5 and approximately 15 years for child mortality. The birth
1955-59, 1960-64,
cohorts are divided into the following periods:
1965-69, and 1970-73 for infant mortality and 1955-59, 1960-64, and
1965-69 for child mortality.
Since demographic and socioeconomic variables are associated with each
other (Hull and Gubhaju, 1984: 1; Frenzen and Hogan, 1982: 398), to
examine the net effect of each variable on infant and child mortality we
have employed a logit-linear model, which is an efficient way to
introduce the necessary statistical controls when
the
dependent
variables are dichotomous and the independent variables are categorical
(Goodman, 1972; Little, 1978: 23-25). Model-fitting in the subsequent
analyses was carried out using a computer package GLIM (Generalized
Linear Interactive Modelling), which was developed by the
Royal
Statistical Society and the Numerical Algorithm Groups (Baker and
Nelder, 1978).
3• Framework for Analysis
The present analysis is carried out with four models. The first model
is used to study the impact of demographic factors on infant and child
mortality, separately for urban and rural areas because the social and
economic situations are still very different between the two areas in
Korea (Korea, 1982). The factors included in this model are sex of the
child (male, female), maternal age (less than 25 years, 25 to 29 years
and 30 years and over), birth order (first birth, second to third
births, fourth and higher order births), and year of birth (1955-59,
1960-64, 1965-69, 1970-73 for infant mortality and 1955-59, 1960-64 and
1965-69 for child mortality) to allow for time trend of infant and child
mortality.
The second model includes the socioeconomic factors: mother's education
(less than 6 years, 6 years and more), father's education (less than 9
years, 9 years and more), father's occupation (professional and clerical
workers, farmers, and others)?- number of rooms used (1, 2, 3 and more
rooms). Year of birth, using the same classification as that used in the
above model, is also added. This model is produced separately for urban
and rural areas.
3
For the third model we select two demographic factors: maternal age* and
education and
birth order; and two socioeconomic factors: mother
— ’-s---------------considered
in
the
earlier
two
models as
number of rooms used which were
Here
determinants
of
infant
and
child
mortality.
the most significant
rural
of
birth
and
analyse
the
model
by
urban
and
again, we add year
areas separately.
In the final model, two other important variables, namely previous birth
interval (less than 24 months, 24 to 35 months, 36 months and more) and
the survival of the previous birth to age one (alive at age one, died
before age one) are incorporated. As we have observed in the preceding
section, previous birth interval and survival of previous birth clearly
affect infant and child mortality. However, there are two limitations
to this model. One is that we have to ignore all first births because
previous birth interval and survival of previous birth are inapplicable
for them. The other is that there are too few cases of deaths of
children whose older sibling died before one year old, to add many
relevant variables for the model.5 Therefore, we will examine the impact
of previous birth interval and survival of previous birth when the
effects of two demographic variables (maternal age and birth order) and
two socioeconomic variables (mother's education and number of rooms
used) are controlled. This model will not consider rural/urban residence
and year of birth.
4. Results
Effects of Demographic Variables on Infant and Child Mortality
Table 1 presents the result of the logit-lihear model of the effects of
demographic variables on infant mortality, for urban and rural areas
separately.^
Only the effect of the sex of the child in urban areas and the effect of
birth order in rural areas are statistically significant (with or
without controls for the effects of the other demographic factors).
Although the gross effect of birth order in urban areas is statistically
significant at five per cent level, it disappears when the other
demographic variables are controlled for.
In urban areas male children have 26 per cent (1.12/.89) higher risk of
infant death than female children which confirms that sex differences in
infant
mortality reflect innate biological differences in infant
liability (Scrimshaw, 1978). However, in rural areas the sex difference
is not significant.
This indicates that, in the Korean rural areas
where the traditional value of children^is still strong (J.H.Cha, 1978),
the better care for sons offsets the biological differences (see
Scrimshaw, 1978) while in urban areas the differences are clear because
of the change in parents' attitudes to their children, smaller family
size, and higher proportion of neonatal deaths under lower infant
mortality.
4
The
net effect of birth order in rural areas is statistically
significant: as expected, the second and third births have a lower risk
of infant death than other births. In the urban areas the same pattern
by birth order applies, but the effect is not significant. Maternal age
is also generally considered as an important determinant of infant
mortality (Gubhaju, 1984; Hull and Gubhaju,
1984). However, in both
urban and rural areas in Korea, the effects of maternal age on infant
mortality are no longer statistically significant, perhaps because of
higher age at first marriage and comparatively early cessation of
childbearing.®
(Tables 1 and 2 about here)
Table 2 presents the effects of demographic factors on child (1-4 years
of age) mortality. This table shows very different results from those
for infant mortality. The effects of the year of birth are significant
in both urban and rural areas. This result may be greatly influenced by
the change in socioeconomic environment,^ as child deaths are more
closely associated with socioeconomic factors than are infant deaths.
Contrary to their effects on infant mortality, the net effects of
maternal age and birth order on child mortality are each statistically
significant in the urban areas, and those of maternal age in the rural
areas. Although the gross effects of maternal age and birth order on
child mortality are not statistically significant in urban areas, when
the effects of other variables are controlled the risk to the youngest
maternal age group (24 years and less) is over four times (1.77/.41) as
high as the risk to the oldest maternal age group and the risk to fourth
or higher order births is over three times (2.15/.68) as high as to
first births. These results may be caused by the strong negative
relationships between number of births and age at marriage, and positive
relationships between age at marriage and educational level. A similar
effect of maternal age occurs in rural areas although the difference in
child mortality level is not as great as in urban areas.
Child mortality is mainly related to exogenous factors, so living
conditions, mother’s experience and attitude to child care may be more
important than maternal age or birth order. Therefore, the risk of child
death is lower for the older mothers who may have better living
conditions and more experience in the care of children; first births
have lower risk of child death because more care may be given to the
first child especially by older mothers. Since these environmental
effects
on
child mortality are stronger in more urbanized and
lower-fertility societies, the values of X^re greater in urban than in
rural areas.
Effects of Socioeconomic Variables on Infant and Child Mortality
Table 3 presents the result of the analysis of the effects of
socioeconomic variables on infant mortality in urban and rural areas.
5
In both areas the gross and net effects of mother's education on infant
mortality are statistically significant.
When the effects of other
variables are controlled the risk of infant mortality among the births
to mothers educated 5 years or less is 63 per cent (1.45/.89) higher in
urban areas and 29 per cent (1.11/.86) higher in rural areas than among
births to mothers educated 6 years and over. On the other hand, although
the effect of father's education without control for the effects of
other
variables is statistically significant in rural areas, no
significant net effects of father's education are found in either urban
This indicates that increasing female education has
or rural areas.
played a major role in the decline of infant mortality, and the
has had more impact upon infant survival than that
education of mothers
----of fathers (see also Caldwell, 1979; Cochrane, 1980 ) .
The effect of number of rooms used, however.
however,
is significant for both
gross effects and net effects in urban areas while it is insignificant
in rural areas. Generally, economic variables may have less effect on
infant mortality. However, if the house is too small for the family to
live in, the crowded circumstances may seriously affect infant mortality
(see Munroe and Munroe, 1971). Although the patterns of the odds ratios
in the 'Net Effect' column in both urban and rural areas are the same,
the situation is more serious in urban areas: the risk of death when the
infant shares only one room with its family in urban areas is 53 per
cent (1.30/.85) to 60 per cent (1.30/.81) higher than if the family
shares two rooms or three or more rooms, respectively.
In the case of child mortality in Table 4 the gross effects of the
socioeconomic variables on child mortality in both urban and rural areas
are statistically significant, except for father's occupation in urban
areas.
However, after the effects of other socioeconomic variables and
the year of birth are controlled for, only the effects of mother's
education in rural areas and number of rooms used in both urban and
rural areas remain statistically significant.
(Tables 3 and 4 about here)
In rural areas the net effect of mother's education still remains a
significant determinant of child mortality. The risk of child death to
mothers educated 5 years or less is 47 per cent (1.16/.79) higher than
to others, which is a steeper gradient than the 29 per cent for rural
infant mortality shown in Table 3. However, mother's education is not
any more a significant factor on child mortality in urban areas. This
result suggests that in a rural society where traditional customs and
values are still strong the roles of mothers are still important for
children's surviving during early childhood.
living
conditions
We noted earlier the effect of the family's
(approximated by the number of rooms used to live in by the family) on
infant mortality, As far as child mortality is concerned, in the urban
areas living conditions exhibit a stronger effect than any of the other
socioeconomic variables used in this model, The risk of death between
exact ages 1 and 5 for children sharing one room with their parents is
over 1.5 to nearly 2.5 times as high as for children living in better
6
<0
conditions. In rural areas this effect on child mortality becomes
statistically significant, although it is not significant for infant
mortality.
Since the number of rooms used is an approximate measure of
the living status of the family it is not surprising that in a society
where living space is limited and houses are in short supply, the effect
of number of rooms is greater for child than for infant mortality.
Effects of Both Demographic and Socioeconomic Variables on Infant and
Child Mortality
In the preceding models we examined the effects of demographic and
i
socioeconomic variables oni infant and child mortality
separately,
However, these variables are often closely associated with each other:
for example, higher-educated women may marry laterand have fewer
children J2, Thus, Table 5 presents the effects of selected demographic and
socioeconomic variables on infant and child mortality. In Table 5 we
displayed only the 'Net Effect' columns by urban and rural areas because
the 'Gross Effect' of each variable is obviously the same as those in
the preceding models.
(Table 5 about here)
The variables with statistically significant net effects on infant
mortality are mother’s education and number of rooms used in urban
areas, and birth order and mother’s education in rural areas. This
again
that
mother’s education is the most important
indicates
areas,
determinant of infant mortality in both urban and rural
socioeconomic
variables
were
relatively
more
important
for
Furthermore,
whereas
demographic
variables
(birth
infant mortality in urban areas
order) retained importance in rural areas.
The net effects on child mortality of each variable, except mother’s
education in urban areas and birth order in rural
areas,
are
statistically significant.
Contrary to the determinants of infant
mortality, demographic variables in the urban areas and socioeconomic
factors in the rural areas are the more important determinants of child
mortality.
These results are basically the same as those obtained in Tables 1 and 3
for infant mortality and Tables 2 and 4 for child mortality. However,
the levels of the effects of birth order, mother's education and number
of rooms used change when the effects of the other set of variables are
controlled. Among them, the change in the effects of mother’s education
is clearer than that of others; the effects of mother’s education on
infant and child mortality in rural areas become more pronounced and the
level of statistical significance changes from 5 to 1 per cent. In
traditional, mother's
rural areas where the society is still more traditional,
education is the most important determinant of infant and child
survival.
7
Effects of Previous Birth Interval and Previous Birth Survival
on Infant and Child Mortality
In Tables 6 and 7, the effects of previous birth interval and previous
birth survival on infant and child mortality are tested with controls
for the effects of selected demographic or socioeconomic variables,
ignoring the year of birth and mother's present residence (urban and
rural). This may have introduced some bias because of the change in
infant and child mortality over time and because of the differentials by
urban and rural areas.
(Tables 6 and 7 about here)
Within these limitations we find strong effects of the length of the
previous birth interval and of previous birth survival on infant
mortality in both the demographic and socioeconomic models. The risk of
infant death among children with previous birth interval of 23 months or
less is still around twice as high as that of other children; death of
the preceding sibling raises the probability of death during infancy for
the younger sibling (index child) more than twice.
There are two interesting results: the effects of birth order on both
infant and child mortality are highly significant (Table 6); and the
effect of the housing conditions of the family on infant mortality is no
longer significant (Table 7). In the earlier analysis the effects of
birth order on infant mortality in the urban areas and on child
mortality in the rural areas were not statistically significant (see
Tables 1 and 2). When the first birth is excluded and the year of birth
and place of mother's residence are ignored, the difference in survival
between earlier and later order births becomes clearer than in earlier
analysis because children are born earlier to younger and more educated
mothers than to older and less educated ones in a society which is in
the process of fertility transition, like Korea. Furthermore, when the
effects of birth spacing and survival of previous sibling are controlled
for, the early childhood mortality differentials by birth order (except
first birth) become much clearer in Korea.
On the other hand, the
effect of the housing conditions of the family (number of rooms used) on
infant mortality was statistically significant in urban areas in earlier
analyses (see Tables 3 and 6), but not in Table 7.
Keeping in mind the limitations of the models, the effect of previous
birth survival on child mortality is not found to be significant in
either of the models. Although previous infant loss increases the risk
that a woman will again lose a child (MacMahon, 1974), it appears that
if the child has survived infancy the risk does not affect its survival
during early childhood (after age 1) in Korea.
Comparison with other
countries (Table 8) supports the view that the effects of previous birth
survival on child mortality may disappear where child mortality has
reached lower levels. In all these instances, in the urban areas, and
in Sri Lankan rural areas, where the mortality levels are relatively low
the effects of previous birth survival are no longer statistically
significant.
8
(Table 8 about here)
The effects of the family living conditions (number of rooms used) on
child mortality are again statistically insignificant. From the result
that even ’Gross effect’ is statistically insignificant, we know that
the effect of number of rooms used diminished when the effects of
urban/rural areas, year of birth and first births on infant and child
mortality were combined or excluded.
5. Summary
In both urban and rural areas, mother’s education is the most important
determinant of infant mortality; maternal age and number of rooms used
are the main determinants of child mortality. Previous birth interval
and survival of the previous birth also significantly affect infant
mortality and previous birth interval child mortality in both urban and
rural areas. For both infant and child mortality, previous birth
interval and number of rooms used are the most general factors in urban
areas; previous birth interval and mother’s education in rural areas.
Infant mortality is also significantly affected by sex of the child and
the number of rooms used in urban areas, and by birth order in rural
areas. Significant determinants of child mortality are birth order in
urban areas but mother’s education in rural areas. Thus,
in summary,
demographic factors are relatively more important for infant mortality
in the rural areas and for child mortality in the urban areas;
in
contrast, socioeconomic factors play a role for infant mortality in
urban areas and for child mortality in rural areas.
Notes
1. This is a substantially revised part of Mortality Transition
in Korea: 1960-1980, submitted for the degree of Ph D.
in
Demography to the Australian National University, 1986.
2. Infant and child mortality in Korea were respectively 57.7
per thousand live births and 45.5 per thousand children aged
one year in 1955-59 birth
cohort,
and
declined
to
respectively 42.5 in 1970-73 and 19.9 in 1965-69 (see Kim,
1986: 43-48).
3. The 1974 KNFS was carried out from September 16 to December
5, 1974 (World Fertility Survey, 1978: 2) and the mid-date of
the period was October 27, 1974. Therefore, the exact periods
of the last cohorts are January 1970 - October 1973 for
infant mortality and January 1965 - October 1969 for child
mortality.
9
4. ’Professional and clerical workers’
include
technical, administrative, managerial, clerical
workers. ’Farmers’ include agricultural, animal
forestry workers, fishermen and hunters; and
sales, services, skilled and unskilled workers.
professional,
and related
husbandry and
’* others*
are
5• Among the total number of second and higher order births, in
the period 1955-73, the number of births in which the
previous sibling died before age one is 748 and the number of
infant deaths among them is only 91. The number of children
aged one whose previous sibling died before age one is 516
for the births in the period 1955-69 and the number of child
deaths among them is only 23.
6. The
results of logit-linear models of the effects of
demographic and socioeconomic variables on infant and child
mortality show the odds ratio for each specific category and
likelihood ratio chi-square test (XL^, degrees of freedom and
corresponding probability value p, for each variable. The
effects of each variable are divided into ’Gross Effect*
and
'Net Effect’. The odds ratio indicates the relative risks of
dying due to being in a specific category.
The likelihood
ratio
chi-square, the degrees of freedom and p value
associated with each variable indicate whether a particular
variable
is
related
to mortality in a statistically
significant manner. The ’Gross Effect’ column indicates the
effect of the specific variable when the effects of other
variables are not controlled for. The net effect of each of
the variables on mortality, controlling for the effects of
other variables, is shown in the ’Net Effect' column.
7. The main values of children in Korea were continuation of the
family, security for aging parents, labour for agricultural
families (J.H.Cha, 1978: 867).
8. The proportions to total births for the extremely high or low
age groups are too small to analyse (6 per cent for less than
20 years and 9 per cent for 35 years and over).
9. The First 5-Year Economic Development Plan was conmmenced in
1962. Thus, the mortality decline could have accelerated
since then.
10. The inverse J-shape of the effects of number of rooms used on
child mortality in urban areas was still observed even after
controlling for the effects of other variables, which result
may be due to some urban characteristics, such as short
supply of housing, smaller family size, economic hardship for
a big house in urban areas (see Kim, 1986:
173-174).
11. The mean age at first marriage of women who married before
age 25 and were 30 to 34 years of age in 1974 was 20.4 years
for the women educated for 1 to 6 years and 22.5 years for
those educated over 12 years (Kim, 1981: 35).
10
12. The mean number of children ever born to all ever-married
women in 1974 was 3.6 for the women educated for 1 to 6 years
and 2.3 children for those educated over 12 years (Kim, 1981:
57).
REFERENCES
ADAMCHAK, Donald J. and Edward G. STOCKWELL
1978
’Trends in the Relationship between Infant Mortality and Socio
economic Status:
1950 and 1970', SOCIOLOGICAL FOCUS, 11 (1):
47-52
ARRIAGA, Eduardo E.
1980
’Direct Estimates of Infant Mortality Differentials from Birth
Histories', WORLD FERTILITY SURVEY CONFERENCE, LONDON, Vol. 2,
Voorburg: International Statistical Institute, pp. 435-466
ARRIAGA, Eduardo E. and Frank HOBBS
'Infant Mortality Differentials in Selected East and South
1982
Asian Countries', MORTALITY IN SOUTH AND EAST ASIA, A REVIEW OF
CHANGING TRENDS AND PATTERNS,
1955-1975, WHO/ESCAP Meeting,
Manila, December 1980, Geneva: WHO, pp. 157-176
BAKER, R. J. and J. A. NELDER
1978
THE GLIM SYSTEM: RELEASE 3, Numerical Algorithms Group, Oxford
BOUVIER, L. F. and J. van der TAK
1976
'Infant Mortality-Progress and Problems’, POPULATION BULLETIN,
31 (1): 3-33
CALDWELL, John C.
1979
'Education as a Factor in Mortality Decline: An Examination of
Nigerian Data', POPULATION STUDIES, 33 (2): 395-413
CALDWELL, John C. and Peter F. MCDONALD
1981
'Influence of Maternal Education on Infant and Child Mortality:
Levels and Causes', INTERNATIONAL
POPULATION
CONFERENCE,
MANILA: SOLICITED PAPERS, Vol. 2, Liege: IUSSP, pp. 79-95
CALDWELL, John C., P. H. REDDY and Pat CALDWELL
1983
'The Social Component of Mortality Decline: An Investigation in
South India Employing Alternative Methodologies', POPULATION
STUDIES, 37 (2): 185-205
CHA, J. H.
1978
'The Value of Children*
(in Korean), in H. Y. Lee and T. H.
Kwon (eds.), KOREAN SOCIETY:
POPULATION AND DEVELOPMENT,
Vol. 3, The Population and Development Studies Center, Seoul:
Seoul National University Press, pp. 849-902
CHOWDHURY, A. K. M. Alauddin
1981
'Infant Deaths, Determinants and Dilemmas (A Cohort Analysis
for Rural Bangladesh)', SCIENTIFIC REPORT, No. 46, ICDDR, Dacca
COCHRANE, Susan H.
1980
THE EFFECT OF EDUCATION ON HEALTH, World Bank Working Paper,
No. 405, Washington: World Bank
EDMONSTON, Barry, William GREENE and Ken SMITH
1981
'Multivariate Analysis of Survival Data: An Appraisal Using
11
Bangladeshi Mortality data*,
1981 PROCEEDINGS OF THE SOCIAL
STATISTICAL SECTION, American Statistical Association, pp. 251255
FEDERICI, Nora and Luciano TERRENATO
’Biological Determinants of Early Life Mortality’,
in S. H.
1980
Preston (ed.), BIOLOGICAL AND SOCIAL ASPECTS OF MORTALITY AND
THE LENGTH OF LIFE, Liege: IUSSP, pp. 331-362
FLEGG, A. T.
'Inequality of Income, Illiteracy and Medical Care as Deter
1982
minants of Infant Mortality in Underdeveloped Countries* ,
POPULATION STUDIES, 36 (3): 441-458
FRENZEN, Paul D. and Dennis P. HOGAN
1982
’The Impact of Class, Education and Health Care on Infant
Mortality in a Developing Society: The case of Rural Thailand’,
DEMOGRAPHY, 19 (3): 391-408
GOODMAN, Leo A.
1972
'A Modified Multiple Regression Approach to the Analysis of
Dichotomous Variables ’,
AMERICAN SOCIOLOGICAL REVIEW,
37
(February): 28-46
GUBHAJU, Bhakta B.
DEMOGRAPHIC AND SOCIAL CORRELATES OF INFANT AND CHILD MORTALITY
1984
IN NEPAL, Unpublished
Ph D Thesis,
Australian
National
University, Canberra
HAINES, Michael R., Roger C. AVERY and Michael A. STRONG
’Differentials in Infant and Child Mortality and their Change
1983
over Time: Guatemala, 1959-1973’, DEMOGRAPHY, 20 (4): 607-621
HOBCRAFT, John N., John W. MCDONALD and Shea O. RUTSTEIN
’Child-Spacing Effects on Infant and Early Child Mortality’,
1983
POPULATION INDEX, 49 (4): 585 ff
HULL, Terence H. and Bhakta B. GUBHAJU
’Multivariate Analysis of Infant and Child Mortality in Java
1984
and Bali', RESEARCH NOTE, No. 27, International Population
Dynamics Program, Australian National University, Canberra
JOHNSON, Gwendolyn Z.
1964
’Health Conditions in Rural and Urban Areas of Developing
Countries', POPULATION STUDIES, 17 (3): 293-309
KIM, Tai-Hun
AGE AT FIRST MARRIAGE AND FERTILITY IN KOREA, Unpublished M.A.
1981
Thesis, Australian National University, Canberra
MORTALITY TRANSITION IN KOREA:
1960-1980, unpublished Ph D.
1986
thesis, Australian National University, Canberra
LITTLE, Roderick J. A.
GENERALIZED LINEAR MODELS FOR CROSS-CLASSIFIED DATA FROM THE
1978
WFS, WFS/Tech. No. 834, World Fertility Survey, International
Statistical Institute
MacMAHON, Brian, Mary Grace KOVAR, and Jacob J. FELDMAN
1972
'Infant Mortality Rates: Socioeconomic Factors’, VITAL AND
HEALTH STATISTICS, Series 22, No. 14, Washington: U.S. Government Printing Office
MUNROE, R. L. and R. H. MUNROE
'Household Density and Infant care in an East African Society',
1971
JOURNAL OF SOCIAL PSYCHOLOGY, 83; 3-13
PARK, Chai Bin and B. T. PARK
INFANT MORTALITY OF KOREA: AN ANALYSIS OF RECENT FERTILITY
1981
SURVEY DATA (in Korean), Seoul: Korean Institute of Population
12
and Health
RINEHART, Ward and Adrienne KOLS
1984
'Healthier Mothers and Children
through Family Planning',
POPULATION REPORTS, Series J, No. 27, Population Information
Program, Johns Hopkins University, Baltimore
RUTSTEIN, Shea Oscar
1983
INFANT AND CHILD MORTALITY: LEVELS, TRENDS AND DEMOGRAPHIC
DIFFERENTIALS, WFS Comparative Studies, no. 24, World Fertility
Survey, International Statistical Institute
RUZICKA, Lado T.
1984
'Birth Spacing and Child Survival: Some Methodological Issues',
RESEARCH NOTE,
No.
11, International Population Dynamics
Program, Australian National University, Canberra
SCHULTZ, T. Paul
1980
'Interpretation of Relations among Mortality, Economics of the
Household, and the Health Environment’, PROCEEDINGS OF THE
MEETING ON SOCIOECONOMIC DETERMINANTS AND CONSEQUENCES OF
MORTALITY, UN/WHO Seminar, Mexico City, June 1979, Geneva: WHO,
pp. 382-422
SCRIMSHAW, Susan C. M.
'Infant Mortality and Behavior in the Regulation of Family Size'
1978
POPULATION AND DEVELOPMENT REVIEW, 4 (3): 383-403
de SWEEMER, Cecile
1984
'The Influence of Child Spacing on Child Survival*, POPULATION
STUDIES, 38 (1): 47-72
SWENSON, Ingrid
1978
'Early Childhood Survivorship Related to the Subsequent Inter
pregnancy Interval and Outcome of the Subsequent Pregnancy',
TROPICAL PEDIATRICS AND ENVIRONMENTAL CHILD HEALTH, 24 (3):
103-106
TRUSSELL, T. James and Charles HAMMERSLOUGH
1983
'A Hazards-Model Analysis of the Covariates of Infant and Child
Mortality in Sri Lanka', DEMOGRAPHY, 20 (1): 1-26
WINIKOFF, Beverly
1983
'The Effects of Birth Spacing on Child and Maternal Health',
STUDIES IN FAMILY PLANNING, 14 (10): 231-245
WORLD FERTILITY SURVEY (WFS)
1974: ,A SUMMARY OF
1978
THE KOREAN NATIONAL FERTILITY SURVEY,
FINDINGS, ’World Fertility Survey, International Statistical
Institute
WRAY, Joe D.
1971
'Population Pressure on Families:
Family Size and Child
Spacing', REPORTS ON POPULATION/FAMILY PLANNING, No. 9, New
York: The Population Council
Table Is
Variables
Gross Effect
Net Effect
Odds
Ratio
Odds
Ratio
2
X
1.04
.86
1.23
1.00
.86
1.23
1
2-3
4+
1.12
1.05
1.02
.86
1.12
1.03
1.02
.88
.2
2.0
1.15
.94
.94
1.14
1.11
.94
.81
.95
.85
1.27
7.5
1.15
1.12
.94
.81
YEAR OF BIRTH
1955-59
1960-64
1965-69
1.43
1.23
.54
1.41
1.34
.63
Grand Mean = .027
Grand mean = .050
LR
.5
9.5
11.9
1.42
1.05
.67
1.27
1.02
.72
5.4
3.2
.90
.81
1.25
1.25
.94
.95
18.4
24.7
14.0
1.31
1.17
.61
2
X
.96
1.05
.96
1.05
.68
.77
2.15
Odds
Ratio
1.0
16.3
19.5
2
X
Net Effect
LR
17.4
3.3
BIRTH ORDER
Odds
Ratio
2.2
1.77
1.03
1.20
1.03
.66
Gross Effect
LR
5.6
MATERNAL AGE
2
X
.88
1.15
.89
1.12
1
2-3
4+
1.04
.80
1.18
Odds
Ratio
1.5
SEX
-24
25-29
30 +
9.4
2
X
Net Effect
LR
Male
Female
1.02
.98
Gross Effect
Odds
Ratio
2
.1
7.6
Variables
X
LR
1.16
.82
1.08
2.2
2.9
YEAR OF BIRTH
Odds
Ratio
9.4
5.6
7.0
BIRTH ORDER
Net Effect
LR
1.08
.90
1.03
.94
1.07
.97
.91
1.03
1.08
-24
25-29
30 +
Rural
Urban
2.6
1.0
1.4
MATERNAL AGE
2
X
1.02
.98
1.12
.89
1.11
.88
Odds
Ratio
4.2
4.4
Male
Female
2
X
Gross Effect
LR
LR
SEX
MODEL IB: LOGIT-LINEAR MODEL OF THE MAIN EFFECTS OF
DEMOGRAPHIC VARIABLES AND YEAR OF BIRTH ON CHILD
MORTALITY BY PLACE OF MOTHER'S RESIDENCE,
KOREA, 1955-69
Rural
Urban
1955-59
1960-64
1965-69
1970-73
Table 2:
MODEL IA: LOGIT-LINEAR MODEL OF THE MAIN EFFECTS OF
DEMOGRAPHIC VARIABLES AND YEAR OF BIRTH ON INFANT
MORTALITY BY PLACE OF MOTHER'S RESIDENCE,
KOREA, 1955-73
1.28
1.24
.67
Grand Mean = .041
Grand Mean = .057
* Significant at 5 per cent level
* Significant at 1 per cent level
2) Sex has 1 degree of freedom; all remaining variables have 2
degrees of freedom each.
Notes : 1)
* Significant at 5 per cent level
* Significant at 1 per cent level
2) Sex has 1 degree of freedom; maternal age and birth order
have 2 degrees of freedom each; year of birth has 3 degrees
of freedom.
Notes : 1)
Source: The 1974 Korean National Fertility Survey
Source: The 1974 Korean National Fertility Survey
Table 3:
Gross Effect
Odds
Rat io
2
X
Net effect
Odds
Ratio
2
X
LR
LR
MOTHER'S EDUCATION
-5 Years
1.41
6+
"
*87
14.9
11.6
FATHER'S EDUCATION
-8 Years
1.12
9+
"
.93
1)
FATHER'S OCCUPATION
.92
Prof. & Cler.
1.01
Farmers
1.02
Others
2.5
YEAR OF BIRTH
1955-59
1960-64
1965-69
1970-73
5.4
.3
1.02
.94
.4
3.5
.87
1.01
1.00
.3
.2
1.04
1.00
.96
.98
1.03
1.07
4.4
7.4
2.7
1.13
1.11
.94
.81
Gross Effect
Odds
Ratio
2
X
1.11
.86
6.0
15.1
Variables
1.11
1.10
.95
.85
2
X
Net Effect
Odds
Ratio
MOTHER'S EDUCATION
-5 Years
1.31
6+
"
.88
4.3
FATHER'S EDUCATION
-8 Years
1.32
9+
"
.79
1)
FATHER'S OCCUPATION
Prof. & Cler.
.88
Farmers
1.73
Others
.95
7.9
Grand Mean = .057
includes professional, technical, adminisNotes : 1) 'Prof. & Cler.'
trative, managerial, clerical & related workers.
2) * Significant at 5 per cent level
♦* Significant at 1 per cent level
3) Mother's and father's education have 1 degree of freedom
each; father's occupation and number of rooms used have 2
degrees of freedom each; year of birth has 3 degrees of
freedom.
6.2
1.03
.91
4.7
8.3
.75
1.10
.77
10.8
6.0
1.69
1.09
.85
1.40
1.08
.87
23.3
25. 1
18.7
Grand Mean - .027
Grand Mean = .050
LR
1.16
.79
.56
1.10
.78
15.5
1.54
1.36
.59
2
X
8.9
3.1
1.55
.62
.96
Odds
ratio
16.7
1.10
.68
1. 10
1.52
.92
Net Effect
LR
2.8
5.0
2
X
1.20
.68
1.24
.87
19.6
1.42
1.24
.54
Odds
Ratio
.1
1.06
.98
12.8
NO. OF ROOMS USED
1.39
1
2
.61
.96
3+
YEAR OF BIRTH
1955-59
1960-64
1965-69
2
X
Gross Effect
LR
LR
LR
12.8
.73
1.04
.96
1.30
.81
.85
Odds
Ratio
Rural
Urban
Net Effect
LR
.6
1.12
1.06
1.02
.86
2
X
1.07
.81
1.06
.89
.98
2.8
1.12
1.05
1.02
.86
Odds
Ratio
.3
.96
1.02
13.8
NO. OF ROOMS USED
1.25
1
.81
2
.84
3+
Gross Effect
1.14
.81
1.45
.89
.6
MODEL IIB: LOGIT-LINEAR MODEL OF THE MAIN EFFECTS OF
SOCIOECONOMIC VARIABLES AND YEAR OF BIRTH ON CHILD
MORTALITY BY PLACE OF MOTHER’S RESIDENCE,
KOREA, 1955-69
Rural
Urban
Variables
Table 4:
MODEL IIA: LOGIT-LINEAR MODEL OF THE MAIN EFFECTS OF
SOCIOECONOMIC VARIABLES AND YEAR OF BIRTH ON INFANT
MORTALITY BY PLACE OF MOTHER’S RESIDENCE,
KOREA, 1955-73
1.32
1.17
.60
1.37
1.23
.64
Grand Mean = .042
Notes : 1) 'Prof. & Cler.' includes professional, technical, administrative, managerial and related workers.
2) * Significant at 5 per cent level
** Significant at 1 per cent level
3) Mother's and father's education have 1 degree of freedom
each; all remaining variables have 2 degrees of freedom each
Source: The 1974 Korean National Fertility Survey
Source: The 1974 Korean National Fertility Survey
<
9
X
A
Table 5:
Table 6:
MODEL III: LOGIT-LINEAR MODEL OF THE NET EFFECTS OF
DEMOGRAPHIC AND SOCIOECONOMIC VARIABLES AND YEAR OF
BIRTH ON INFANT AND CHILD MORTALITY BY PLACE OF
MOTHER'S RESIDENCE, KOREA, 1955-73
Odds
Ratio
2
X
Odds
Ratio
BIRTH ORDER
1
2-3
4+
1.06
.87
1.20
Odds
Ratio
2
X
10.6
1.42
1.07
.66
16.9
NO. OF ROOMS USED
1.32
1
.81
2
.83
3+
YEAR OF BIRTH
1955-59
1960-64
1965-69
1970-73
2.7
1.14
1.05
1.02
.86
15.7
18.2
4.0
1.09
1.10
.96
.85
6.9
1.46
1.09
.87
1.57
.63
.94
1.04
1.03
.97
1.49
1.42
.58
.0
2
X
Net Effect
Odds
Ratio
2
X
LR
LR
11.6
24.5
1.69
1.20
.61
1.25
1.10
.75
11.7
14.6
.84
1.21
.85
1.16
20.0
2.2
.72
1.43
.91
1.09
15.3
15.9
.3
Odds
Ratio
LR
.99
.99
1.01
.96
.96
1.06
BIRTH ORDER
1.24
.74
1.16
.95
1.14
.84
2
X
Gross Effect
4.7
.9
8.5
5.3
Odds
Ratio
1.4
MATERNAL AGE
4+
MOTHER'S EDUCATION
-5 Years
1.31
6+
H
.92
2
X
Net Effect
LR
-24
25-29
30+
.95
.82
1.21
.69
.79
2.05
Gross Effect
LR
14.3
8.5
1.08
.81
1.15
Odds
Ratio
18.3
1.71
1.10
.39
1.16
.95
.92
4.9
2
X
LR
2.0
2.0
.90
1.10
.98
Odds
Ratio
LR
LR
MATERNAL AGE
-24
25-29
30+
2
X
Variables
Rural
Urban
Rural
Urban
Child
Infant
Child
Infant
Variables
MODEL IVA: LOGIT-LINEAR MODEL OF THE MAIN 'EFFECTS OF
MATERNAL AGE, BIRTH ORDER, PREVIOUS BIRTH INTERVAL AND
PREVIOUS BIRTH SURVIVAL ON INFANT AND CHILD MORTALITY,
KOREA, 1955-73 1),2)
1.25
1.24
.68
PREVIOUS BIRTH
INTERVAL
-23 Months
24-35
36+
PREVIOUS BIRTH
SURVIVAL
Alive at Age 1
Dead before
Age 1
55.9
78.4
63.4
.90
2.61
.0
2.0
32.0
.95
2.05
1.50
.80
1.07
1.58
.80
.93
1.73
.88
.81
1.74
.81
.76
17.0
22.0
.98
1.35
.99
1.04
Grand Mean = .035
GRAND MEANS
.050
.057
.026
.041
* Significant at 5 per cent level
* Significant at 1 per cent level
2) Mother’s education has 1 degree of freedom; year of birth
for infant mortality has 3 degrees of freedom; all remaining
variables have 2 degrees of freedom each.
Notes : 1)
Notes :
1) The first order births are excluded.
2) Corresponding years are 1955-73
for
infant mortality and
1955-69 for child mortality.
3) * Significant at 5 per cent level
** Significant at 1 per cent level
4) Birth order and previous birth survival have 1 degree of
freedom each; maternal age and previous birth interval have
2 degrees of freedom each.
Source: The 1974 Korean National Fertility Survey
Source: The 1974 Korean National Fertility Survey
Table 7:
Table 8:
MODEL IVB: LOGIT-LINEAR MODEL OF THE MAIN EFFECTS
OF MOTHER'S EDUCATION, NUMBER OF ROOMS USED, PREVIOUS
BIRTH INTERVAL AND PREVIOUS BIRTH SURVIVAL ON INFANT
AND CHILD MORTALITY, KOREA, 1955-73 1),2)
Gross Effect
Odds
Ratio
2
X
Net Effect
Net Effect
2
Odds
Ratio
Odds
Ratio
X
2
X
INDONESIA
67
83
SRI LANKA
21
26
1.33
.73
1.25
.73
1.25
.82
4 Demographic
2 Socioeconomic
7)
2 Demographic
1.31
.96
.92
1.25
.97
.92
69
89
2 Socioeconomic
5)
4 Demographic
NEPAL
110
2 Socioeconomic
5)
4 Demographic
78.4
.96
2.02
1.70
.83
, .94
1.58
.80
.93
.98
1.35
Notes
.0
1.9
31.0
63.5
Grand Mean = .053
/
.99
1.04
Grand Mean = .035
Notes : 1) The first order births are excluded.
2) Corresponding years are 1955-73 for infant mortality and
1955-69 for child mortality.
3) * Significant at 5 per cent level
* Significant at 1 per cent level
4) Mother's education and previous birth survival have 1 degree
of freedom each; number of rooms used and previous birth
interval have 2 degrees of freedom each.
Source: The 1974 Korean National Fertility Survey
*•
.2
8.3
NA
1.6
1.2
.1
NA
33.7
.6
34.6
NA
8.2
NA
6.9
23.4
21.9
57.2
1.73
.88
.81
1.74
.81
.76
.90
2.62
7.2
6)
2 Socioeconomic
PREVIOUS BIRTH
SURVIVAL
Alive at Age 1
Dead before
Age 1
.0
U
6)
5.0
3.8
4.6
1.18
.96
.95
1.15
.97
.95
PREVIOUS BIRTH
INTERVAL
-23 Months
24-35
36+
Rural
6)
BANGLADESH
1
2
3+
Urban
6)
3.6
NO. OF ROOMS USED
Likelihood Ratio X
X
23.5
18.7
24.7
2
LR
LR
LR
17.1
1.18
.84
-5 Years
6+
Gross Effect
2 4)
3)
Controlled
Variables
5)
Odds
Ratio
LR
MOTHER'S EDUCATION
q 2)
4 1
Countries
Child
Infant
LIKELIHOOD RATIO CHI-SQUARE OF NET EFFECT OF PREVIOUS
BIRTH SURVIVAL ON CHILD (AGE 1 TO 5) MORTALITY,
SELECTED ASIAN COUNTRIES, 1962-71 1)
: 1) Model used is logit-linear model.
2) The probabilities of dying between exact ages one and five
years (per 1000 children at age one).
3) All models include previous birth interval as a control
variable.
4) Each has 1 degree of freedom.
age. sex of the child and birth
5) Birth order, maternal age,
cohort.
6) Mother's education and father's education.
Birth order and maternal age.
NA: Not available
Significant at 1 per cent level
Sources: World Fertility Survey
Nepal : Gubhaju, 1984
Others: Unpublished data in the Department of Demography,
ANU, Canberra
7
Research Note on
CHILD
SURVIVAL
Number
lies
Date
24 July 1986
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
4
METHODS OF MORTALITY MEASUREMENT AND ANALYSIS:
A REVIEW
L.T. Ruzicka
Departmental Visitor,
Child Survival Project,
International Population Dynamics Program,
Department of Demography,
The Australian National University
Child Survival Research Notes are brief discussions of
issues of current relevance to researchers and policy
makers concerned with problems of high infant and child
mortality in the world. The International Population
Dynamics Program, Department of Demography, The Australian
National University, distributes these notes with their
regular Bibliographic Circular. Production of the Child
Survival Research Notes is made possible through a grant
from the Ford Foundation (840-0893). Responsibility for
)
the content of Child Survival Research Notes rests
//' with the author(s) alone, and not the above-listed
organisations.
Note:
L'Bf>
Research Note No. 11CS
Methods of Mortality Measurement and Analysis:
A Review
1,
In troduc t i on
Conventional1y, measures of mortality are derived from death registration
statistics in combination with estimates of the appropriate population at
risk. In most countries, however, registration statistics are not available
at all or are incomplete and deficient.
In this note attention will be drawn
to the approaches developed for the estimation of mortality in the situation
of incomplete vital registration, with emphasis on estimation of infant and
child mortali ty.
In many demographic investigations the chief concern has been the
measurement of national, overall, levels, and trends in mortality.
However,
more recently, attention started to focus on the exp 1 anati on of levels and
trends and on the investigation of different i als in mortality levels among
various sub-groups of the national population.
In this note some
observations will be made on this aspect of measurement of mortality.
Finally, investigations of differential mortality lead to attempts for
explanation, that is the identification of the factors that may be responsible
for the observed differences in levels and trends of mortality among
population sub-groups.
This approach calls for development of conceptual
models in which the path and direction of the causes and effects is posited
and to application of multivariate statistical techniques of analysis.
This
note will draw attention to selected applications of such models in recent
years.
2.
Approaches to incomplete vital req i str-at i on [ 1J
In principle, there are several approaches to the estimation of mortality
i n the situation o-f incomplete vital reg i str at i on : (i) to obtain informati on
on the number of deaths (by age and sex) during a fixed period of time through
special questions in censuses or surveys; (ii) to correct incomplete returns
of the vital registration by application of a correction factor that may be
derived individually for each sex and for major age groups; (iii) to implement
an intensive collection of vital statistics in a selected area or on a sample
bas is.
All three approaches may be labelled "direct11 methods as they attempt to
obtain unbiased data o-f the conventional type.
It must also be noted that,
to be applicable, they require a full count of the population at risk or at
least an estimation of the size and age structure of that population.
<i) The question on the number of deaths during a period preceding a
census or a survey may provide adequate data for mortality analysis if the
reference period is clearly identified in the mind of the respondent and if
there is little chance that, for cultural or other reasons, respondents will
avoid reporting deaths of their children or close re 1 atives . [2]
Further
more, in such single-round investigation, certain groups of respondents (such
as better educated persons) may provide more accurate and complete information
1
than others.
This may bias estimates of differential mortality among variouspopulation sub-groups.
An improvement upon the completeness of death
reporting is the introduction of multi-round household surveys in which the
number of deaths (by sex and age) that occurred between two subsequent visits
is obtained.
On the other hand, multi-round surveys require careful matching
of the population coverage from one round to another.
Migration of
population, both in and out, may lead to errors in estimation of the size and
structure of the base population as well as in the reported deaths.
In some societies death o-F the household head often leads to
disintegration and migration of the surviving household members,
Such
mobility may lead to omissions of these deaths,
In-Fant deaths may also be
under-reported in the target population in societies where women move away to
their paternal household to give birth.
(ii) Corrections of existing but incomplete death registration data make
use of the sex and age distributions of recorded deaths.
Several such
methods were developed during the last decade.£31
Some of them assume that
above a certain age, say, five years, the rate of under-registration does not
vary significant1y.
Others also assume that the target population has a
stable age structure,
Many of these procedures are heavily influenced by the
data for older ages.
If age at death is misreported for older persons the
methods become less reliable.
Similarly, they can produce erroneous- results
in situations in which mortality has been declining.
Finally, some of the
methods (notably the Courbage-Fargues method) are sensitive to the choice of
reference model life table.
The user must be familiar with the assumptions
of the method he or she selects and should examine carefully to what extent
these assumptions are met in reality, to avoid grossly biased estimates.
(iii) To obtain better vital statistics, intensive coverage of a sample
population has been introduced in some countries.
Demographic Surveillance
System developed by the former Cholera Research Laboratory (now International
Centre for Diarrhoeal Disease Research) in Dhaka, Bangladesh, has been in
existence since the mid-1960s in a rural area of Bangladesh (Matlab thana in
Comilla district) and later introduced in another area (Teknaf thana) in the
southern part of the country.
This is a three-tier system combining
reporting by village-based workers with regular rounds of field assistants
(who fill out the registration forms for births, deaths, marriages, divorces,
and migrations) and less frequent regular checks of completeness by
supervisons.[4]
In most instances the death and other vital statistics are obtained in
duai-record system combining continuous registration and independent,
multi-round surveys.
Based on the assumption of independence between the two
systems, an estimate of undercount in each collection is obtained by matching
the records, isolating those missed in either collection and then applying the
Chandrasekaran-Deming formula,£53 various modifications to this system have
been developed and are in use in India, Bangladesh and Pakistan, for instance.
The matching procedure sometimes creates a problem, as the number of
events missed in one collection and recorded in the other one, and vi c e ve r s a,
determines the correction factors for events- deemed to be missed in both
collection systems.
Inadequate matching procedure may declare less events as
matched and, hence, over-estimate the corrections; or it may declare as
matching some events that are, in reality, different and thus bias the
resulting estimates downward.[6]
2
The second, and probably nowadays the most frequently used type of
a p p r o a c h e s to estimation of m o r t a1 i t y levels, is through i n d i r e c t a p p r o a c h es.
They seek to obtain indices of mortality through the use of demographic models
and conversion of the calculated measures into conventional life table
parameters.
A review of these approaches is in the United Nations manual
X.C7]
The oldest form of indirect estimation is the inter-censal survival
method.
The numbers alive in a given age X at one census are compared with
the number still living at age X + n at the subsequent census taken n years
later.
There have been many modifications to this general principle
underlying the method, dictated by the necessity to overcome two typical
defects of population censuses: under-enumeration and age misreporting.
None of these attempts has been fully successful, however.
Similar rather limited success has been met with methods o-f mortality
estimation that are based on the assumption that the enumerated population
a stable or quasi-stable one.
j s
The most commonly used methods for estimation of infant and childhood
mortality are now those methods which make use of the reported- number of
children ever born and number of children still riving by age of the
responding women (or, alternatively, by the respondent s duration of
marriage).
It is important to note the assumptions on which these methods
are de~ived:C8j
the mortality schedule is assumed to be constant over
the survival period of children ever born to women i n
a g i ven age categoryj
the fertility indices implied in the method should refe rto the cohort experience of women t n a given age category,
or the fertility schedule should have been constant i n
the past:
the model which predicts the time distribution o-f births
on the basis o-f the indices o-f -fertility must be adequate.
Needless to say, the accuracy of the estimates of mortality depends on
the er-rors in the input data: failure to report children who died, and
misstatement of age or marriage duration of the respondent are the most c ornmor
errors.
There are a number o-F variants o-f the basic method as it was originalb'
deuehoped by h’. Brass in the mid 1960s.
Ore o4 the most use-ful ones by
Preston and Palionir*?] makes use o-f data or the age distribution o-f the
sur ■■■: ■ i ng chi: dr er ,
Such data are usually obtained -from detailed pregnancy
or b;rth history of respondents or -from tabulations
--- -— — o-f
- • own-children,
I-f
these data ar‘e available and age of children is correctly stated, this method
■ b 1 i t e r a t e s the necessity to adopt an assumption on -Fertility distribution,
and hence :s capable of producing more accurate estimates of child mortal it?.
Another approach to indirect estimation of child mortality uses the
survival status of the most recent birth to estimate current or very recent
eve; s of .-r + ant mortality.
An approximate indicator of survival to the age
of two years, q(2), can be obtained if, at the registration of a birth, the”
survival status of the previously born child is ascertained.
3
The indirect methods based on the reported number of children ever born
and children surviving are suitable to estimate mortality between birth and
ages 2, 3 or 5 years.
They provide a poor estimate of infant mortality,
because that index, (1), is based on births to women aged 15-19 years,
In
almost all populations the number of births to this age category will be
small, and hence the chances of random error are large.
In addition, almost
all these births will be first births which are well known to have higher
probability of infant death than higher order births,
Thus, q(l) estimates
based on these data would be biased upward.
Estimates o-F survival -From birth to age 10 and 15 are also possible but
are considered unreliable because o-F the likely reporting errors - omissions
o-F births and child deaths by older women aged 35-39 and 40-44 years.
Estimates o-F adult mortality are more di-F-ficult to obtain by indirect
methods.
Although at least two approaches have been experimented with - one
cased on survival o-F parents and the other on survival c-F spouse - neither is
quite sat i s-Fac tory.
Parents'’ survival obviously re-Fers only to the
experience o-F persons with surviving children.
Moreover,
Mor e over as each surviving
child c-F a mother reports in respect o-F the same woman, t h e w om e n vj i t h m a n y
surviving children will be over-represented,
Theoretically at least methods
based on reports o-F the survival o-F a personas
person-'s -First spouse are likely to
provide better- estimates o-F adult mortality than those based on parents'*
survival.
For one, the time reference period -For the survival o-F spouse will
be generally shorter; in addition, this approach allows the choice between
respondert's age or marriage duration and, hence, the analyst may select the
variant which is likely to yield more reliable results.
There are, though,
minor selection problems? single persons are obviously left out and so are
couples where both have died.
Problems of reporting are also enhanced in the
societies where a large proportion of first marriages is divorced.
To conclude this section, one major drawback o-F the indirect methods
should be noted.
All o-F them are based on data which do not re-Fer to a
particular period o-F time, but rather re-Flect the status o-F the respondent
with respect to some char ac ter i st i cs - -For instance, whether her mother, or
child, or parent is still living.
In retrospective surveys, time related
:n-Formation is usually de-Fective while current status in-Formation may not be.
Direct approaches, in contrast, are time dependent - the data re-Fer to a
particular period,
However, in this instance the major problems arise -from
the completeness o-F the input data.
-T ob • ems in th; -.easuremert
dH-Ferent i a 1 motai ty
D ; t -F e r e n c e s in the level of mortality among various subgroups of the
population are, in most instances, difficult to capture by direct methods of
m o r t a 1 it y m e -a s u r e m e n t >
As long as we deal with comparative1y large
population aggregates (mal es/f emal es; population o4’ major regional
sub-divisions ox a given country; mortality by major ethnic groups;
urban/r ur-al) t may be possible to use the conventional approach: caJ cui at i on
of age-sp e : ■: c death rates, of life tables, of age-sex adjusted
•. standard i zed; indices, and so on,
^he necessary condition is that the
characteristic differentiating various population subgroups must be contained
n both sources of data, that is, in the death registration and census
er,'jmerat ■ or., and must be defined in the same terms.
Secondly, there must be
no or negligible transfe-'*- from one category o* a given variable into another
category during the relevant period:
4
-For instance, urban-rural d i-F-Feren t i al s in mortality would
be biased i-F a proportion o-F rural dwellers-' deaths occurred
in urban hospitals and i-F the deaths were shown as occurring
in the urban areas in the death registration statistics.
Certain important dimensions o-F d i-F-f er en t i al mortality cannot be measured
by conventional methods at all -For the simple reason that the reg i s trat i on o-F
deaths does not contain the relevant i n-Format i on (although such
characteristics may be available in the census data).
The United Nations
recommendat i ons -For the content o-F death registration include - in the -First
priority list - only age, sex, and place of usual residence.
Although the
second priority list recommends collection of information on marital status,
occupation, industry, education and number of children ever born, most
countries do not obtain this information.
Furthermore, some of these
characteristics even if collected may be coded and tabulated differently in
the death registration and in the census (this is often the case with respect
to occupation and industry).
In countries where complete death registration exists, elaborate
techniques have been developed to overcome these problems:[10] on balance,
however, they are unlikely to be applicable in the countries where vital
registration is adequate.
For most instances we have to derive indices o-F d i -F-Feren t i al mortality
There are some advantages to this
■From data obtained by indirect methods,
procedure over the conventional approach: namely, the data on population at
risk and on death come -from the same source
census or survey - and hence the
problem ot comparabiIity mentioned above does not ar i se .
One major problem is to what extent the assumptions underlying the
estimation procedures in the indirect approach are violated once we move -From
national aggregates to sub-populations; tor instance, the age patterns o-F
•Fertility may be very di-F-Ferent -For educated and -For illiterate women; -For
rural and urban women; and -For various ethnic groups in the population.
Some
socio-economic characteristics remain invariable over time (ethnicity) others
may change over time (residence, occupation).
In retrospective surveys we
have i n-Format i on about the current status o-F the household or about attributes
o-F the individuals.
And these characteristics may be di-F-Ferent -From those 15
or 20 years earlier when some o-F the children whose survival indices we
measure were born.
For these reasons as well as -For other concerns (largely
about recall lapse a-F-Fecting the completeness o-F reporting) analysts o-Ften
select a limited period o-F time preceding the survey - say, -Five to ten years
- -For analysis o-F d i-F-Ferences in mortality among various social strata.
Biases arising -From selection in retrospective surveys pose another
problem in d i -F-f eren t i al mortality analysis.
This a-f-Fects, in particular,
indirect estimation o-F adult mortality.
The methods will yield unreliable
results i -F there is any relationship between socio-economic status and marital
status or between such status and -Fertility,
Also, methods based on parental
survival may not be usable i-f the surviving child is expected to report on
characteristics o-F the mother which he/she may not remember or is unlikely to
k n ow.
The spouse survival method may be di-F-Ficult to apply in the case o-F such
char ac ter i st i cs o-F the deceased husband that the widow may not know.
Also,
i-f the widow - as is the case in some cultures - now lives in her son's or
daughter's household, the status and char ac ter i st i cs o-f this household may be
5
very different from those where she lived while her husband was alive - and
obviously it is the latter we are interested in.
To sum up, many indirect methods use assumptions that are difficult to
Justify for other than national estimates, and require data that are
impossible or difficult to obtain for population sub-groups.
This may be one
of the reasons why analysis of differences in mortality has been given
inadequate attention in the past.
4,
Search for explanations of mortality trends and differentials
In many instances the approach to the study of differential mortality has
been through ecological correlation.
For each geographical or administrative
area of the country, level of mortality is calculated (using age-sex
standardized death rate, infant or child mortality rate, or life expectation
at birth as alternative dependent variables) and then the index is correlated
with the socio-economic profile of that area - as, for instance, expressed by
percentage of urban population, adult literacy ratio, gross regional product
per capita, percentage of adults working in agricu1ture, doctor/popu1 ation
ratio, and so on.
With the availability o-F individual charac ter i st i cs of the respondents ii n
demographic surveys, the focus turned more and more towards the study of
mortality differentials between aggregates of individuals sharing the same
attribute: women with no formal
-Formal education as against those with primary or
higher education; urban v_s, rural; women employed outside their home as
against those attending to household chores only, and so on.
Many of the attributes of individuals are inter-dependent: for instance,
education and occupation; occupation and income; housing and- income,
--- , 11c .
Statistical models of multivariate analysis have been used extensively in this
ar e a of i n ve s t i ga t i on .
However, very often the examination of the
assumptions of the statistical model is inadequate and an inexperienced
analyst may select the inappropriate model - sometimes merely because a
package program exists on the computer to which he or she has an easy access.
In principle, two types o-F models may be discerned:
interact!ve.
additive and
For example, a study of de terminants of sex differentials in infant and
child mortality may consider socio-economic status of the parents
approx imated by mother's education, for instance) and nutritional status of
the children as likely factors.
An additive model implies that the economic
and nutritional status indicators mediate the effect of sex biases entirely.
Sex of child
(female = 1)
mortality
measure
nutritional
adversity
economic
(malnourished
adversity
(low SES = 1)
+
6
1)
L 882
SCHOOL
472>
\ 614
. 394
POVERTY
. 820
. 453
NEONATAL
/ 573
BIRTH
ATTENDANT
667
642
. 925
. 663
LITERACY
\ . 464
INFANT
MORTALITY
. 579
VACCINATION \
. 564
. 426
. 688
MEDICAL
-.395 >
334
POSTNEONATAL
;76i
FIGURE 2:
A COMPOSITE MODEL FOR THE DETERMINANTS
OF INFANT MORTALITY
However, if sex preference has a residual effect in the absence of economic
and nutritional adversity, an interactive model may be more appropriate.
A female child may be hypothesized to have better chances to survive on
biological grounds (“sign in the above figure).
However,, in the presence of
However
economic or nutritional adversity or both this biological advantage may be
reversed.
.
In addition, it is assumed in this model that girls are more
subject to nutritional disadvantage than boys (due to maldTstribution of food
within the family, for instance).
The model always represents a simp1ification of reality.
However, even
within such limitation the statistical model may or may not take into
consideration all possible interactions among the independent variables.
A
saturated model would be such a case.
Saturated models may reveal
synergistic conditions where they exist; Jointly present conditions have
additional effect on the risk under study.[11]
The factors affecting the risk of infant and child mortality are
sometimes considered at two levels of analysis: communal variables, such as
existence of a health centre in the village, tube well with safe water to
drink, and other communal amenities, affect the environment in which the
children live and the ecology of the environment and its related risks of
morbidity and mortality.
The characteristics of individuals and f am i1 i e s
provide the second tier of analysis.
A conceptual model was developed recently by Jain[12] for the analysis of
infant mortality rates by State in India,
In his scheme, the content of
which was largely determined by the available data there were actually three
levels of variables:
cummin i ty
- general development of the village approximated
to by the existence of a high school;
- availability of health facilities in the
v i11 age;
househcld
- economic status approximated by an index of
pover ty;
- female literacy;
individual
- medical care at the birth of the child;
- percentage of children immunized by triple
vacci nat ion.
The model is s h ow n in Figure 2 along with the results of multivariate analysis
(path analysis).
The six factors used in the analysis explained about 77 per cent of
regional variation in infant mortality,
The community level factors
transmitted their effect through household and individual factors,
Among the
household variables, poverty appeared to have a direct effect on infant
mortality, while the effect of female literacy was transmitted through the
individual-I eve 1 variables,
In particular, adult female 1 iteracy affected
neonatal mortality through an increase in the proportion of births attended to
by trained medical personnel.
7
The two most important variables in the determination o+ post-neonatal
mortality were availability in the village o-F a health care facility and the
proportion o-F in-fants immunized by triple antigen,
Interestingly enough, the
negative correlation o-F -Female literacy with medical care at birth, and with
curative and preventive medical care in the post-neonatal period accounts -for
about 86 per cent o-F the total e-F-Fect o-F -Female 1 i teracy on in-Fant mortality.
In recent years a statistical method that has been increasingly used in
demographic analysis is a propertional-hazards model.
It is a non-linear
regression technique well suited in situations where observations are censored
as is the case in many retrospective and prospective surveys.
The general
surveys.
formula is
Kt; z) = e x p < z b,) 1 r
where Kt;z) denotes the instantaneous probability o-F -Failing at time t o-F an
individual with a vector o-F initial co-variates z, b, is a vector o-F unknown
regression parameters, and l0□ (t) is a hazard -Function giving the baseline
probability o-F -failing at each distinct duration at which an end point was
observed.[ 13]
Normally, the e-f-fects o-F the initial co-variates on an
individual's -Failure probabilities are assumed to be constant over the
duration o-F observat i on .
However, it is possible to relax this assumption
and allow in a modified model -For time trends,
It is also possible to allow
in the specification of the model for interactions between co-variates.
In several studies based on the World Fertility Survey data , 1og-1i near
models were used and extensive statistical studies were conducted on the
appl i cabi 1 i ty o-F these models to WFS data and demographic problems in
general.[141
A simple example is o-F D jj being the expected number o-F deaths
in a cell o-F a two-dimensional table, and N j j the number ot survivors in that
category at the beginning o-F the period:
In (D{j ) = In (Njj) + a + lj +
CJ
In this additive log-linear model
and c; are parameters corresponding to a
dummy variable -For presence in Row £ and column £, and a is a constant.
Such
models as the one described above have been used extensively in WFS data
analyses.[15]
8
I
FOOTNOTES
[IJ
This section is based on P.F. Macdonald, "The measurement of differential
mortality in the absence of complete death registration statistics", in
, Mortality in South and East Asia: a review cf Changing Trends and
Patterns, 1950-1975: pp. 511-529. Manila, 1982.
[23
About the e-f-fect o-f under-report ing o-f deaths in the.? census enumeration
see Kwok Kwan Kit, "Trends and d ii -f-f
i
-f -f eren t i al s in
infant mortality in
Mai ays i a", ’n
Mortality in South and East Asia; 267-285. Manila,
1982.
[31
Brass, W., Methods -For Estimating Fertility and Mortality from Limited
and Defective Data. CCarolina Population Center, University o-f North
Caroli na, Chapel Hill,, 1975.
Courbage, Y. and P. Fargues, "A method for deriving mortality estimates
from incomplete vital statistics", Populat i on Studies, 1979, 33(1 ):
165-180.
United Nations, Manual X?____________
Indirect Techniques tor Demographic
Estimation, New York, 1983 (Chapter U).
[4]
D"'Souza, S., “Small-area intensive studies -for understanding mortality
and mor bi d it y pr ocesses", In United Nations, Data Bases for Mortality
Measurement. New York, 1984, pp. 146-158.
”
[51
Chandrasekaran, C, and W.E. Deming, "On a method o-f estimating birth and
death rates and the extent o-f registration'’. Journal o-f the American
Statistical Association, 1949, 44(245). 101-115.
[6J
See, for instance, P. Padmanabha, "Use o-f sample r eg i strat i on system -for
studying levels, trends and differentials in mortality". In Un i ted
Nat ions, Data Bases for Mortality Measurement: pp. 54-65. New York,
1984.
C7j
Op. c i t» in +’oat note [3J.
[8]
Op. c i t. in -footnote £13.
[?]
Preston, S.H. and A. Palloni, " Fine-tuning Brass-type mortality estimates
with data on ages of surviving children", Population Bulletin o-f the
Un i ted Nat i ons, No. 10-1977. (E.78.XIII.6): 72-87.
[103 Lynge, E., "Experiences in estimating differentials in mortality in
developed countries". In United Nations, Data Bases for Mortal i ty
Measurement: pp. 117-126. New York, 1984.
£113 This example is taken from Bairagi, R., M.K. Chowdhury and J.F. Philipps,
"A multivariate logistic regression analysis of childhood survival". In
ISTP, Infant and Child Mortality in Bangladesh: 97-110. Dhaka, 1982.
[123 Jain, A . K, , "Determinants o-f regional variations in infant mortality in
r u r a 1 India", The Population Council, New York, Working Paper No.
20/June 19S4.
9
[LSI Santow, G. and M. Bracher, "Child death and birth intervals in ‘Java”,
Population Studies, 1984, 38(2): 241-254.
’
Kay, R., ’’Proportional hazard regression models and the analysis o-f
censored survival data”, Applied Statistics. 1977, 16: 231-234.
[143 Little, R.J.A. ’ l-'terier'5^ i zed
Linear Models -for- Cross-Cl ass i-F i ed Data -From
, World Fertility Survey, London, No. 5/TECH.834, October 1978.
PullurTi, T.W., Standardization , World Fertility Survey, No. 3/TECH.597,
August 1978,
[15] For instance, Rutstein, S.O., Inf’ant and Child Mortality; Levels, irends
and Demographic D i-F-f eren t i al s. WES Comparative Studies No. 24, WES,
London, 1983,
Bobera-ft, J.N., J,W. McDonald and S.O. Rutstein, "Socio-economic -factors
in in-fant and child mortality: a cross-nat i oral comparison’', Pop u 1 a t i on
Stud i es (-for thcomi ng, 1984),
; ■ ?
if
10
.
Research Note on
CHILD
SURVIVAL
Number
Date
13CS
26 November 1986
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
SEX DIFFERENTIALS IN CHILD MORTALITY,
AUSTRALIA, 1909-1984
Terence H. Hull
International Population Dynamics Program
Department of Demography
The Australian National University
Note:
u f
• >
A AJO
Child Survival Research Notes are brief discussions of
issues of current relevance to researchers and policy
makers concerned with problems of high infant and child
mortality in the world. The International Population
Dynamics Program, Department of Demography, The Australian
National University, distributes these notes with their
regular Bibliographic Circular. Production of the Child
Survival Research Notes is made possible through a grant
from the Ford Foundation (840-0893). Responsibility for
the content of Child Survival Research Notes rests with
the author(s) alone, and not the above-listed organisations.
SEX DIFFERENTIALS IN CHILD MORTALITY,
AUSTRALIA, 1907-1984
Terence H. Hull
Coor d i n at or
International Population Dynamics Program
Demogr ap h y De p ar tme n t
Australian National University
INTRODUCTION
Austral i an 1 i +e tables c o n s i s t e n 11 y sh ow that women
live longer than men (1).
Despite the fact that males a r e
about 5 per cent more numerous than females at birth, the
lower mortality of females over the life eye 1e me an s t h a t a t
retirement wome
womenn outnumber men by a factor of over 115 to
1 00 . Their predominance rises considerably through old age.
This paper explores a critical time in the development
of sex differentials in mortality by focussing on deaths
occurring between 1 and 10 years of age.
The data used in
the study are from Austral ia and cover the period from 1907
to 1984.
It is found that over the period in question there
has been a great decline in childrerr's death rates. This hasbeen accompanied by a major rise in the sex ratio o+ deaths,
to the extent that today boys of the age group are between
30 and SOX more likely to die each year than are girls.
The reasons for this
th i s large differential centre largely
on differences in exposure to risks of morbidity and
mor t a 1 i t >- due to accidents. The paper speculates on how
these differences arise in Australia, and finally explores
some of the policy implications of the findings.
THE GROWTH OF INEQUALITY,
1707-1904.
In 1907-09, the first few years for which a c c u r a. t e
natioral data are avai1abie, 8111
chi 1 dren aged 1 to 4 die d
in Aust ra1 i a. T h e sex ratio of deaths was 110.7 which i 8
slightly higher than the suerage sex ratio of births.
□ '•..■erthe next twenty years (see Table 1) the sex ratios drifted
up gradually to reach a 1 e v e 1 of about 120 and thereafter : t
•' particular years, but
occasionally reached 1 30 or mo'- e
U S t as often dropped back to the 120 mark.
S i nee 1970 the
Since
a n d in 1970,
sex ratios h 3 u e a e r a g e d o u e r 1 30 each year, and
1 976, and 1980 haue been ouer 1 40 , In. 1984 the ratio was-
-i -
152. Thus in the last 60 years or so the mortality levels of
ery young girls and boys have changed in quite different
ways to produce a con
contemporary
temperary situation of great inequality
i n mor tai i ty at young ages.
IJ
(TABLE 1 ABOUT HERE)
For the age group 5 to 9 years the pattern is roughly
s i m i 1 ar , bu t the sex ratios start out at around 112 on
average in 1907-09, and rise faster to levels con si s t e n 11 y
above 140 after the late 1940's.
In 1975-79 the sex ratio
of deaths among 5-9 year olds was over 180.
It must be remembered that the period in question was
one of remarkable decl ines i n mor tali ty for both sexes.
The
over 8000 deaths to the 1-4 year olds in the three year
period from 1907-9 can be compared with just over 2000
deaths in the five y e a r s f r om 1980-84, drawn from a much
1 anger base population,
The rates of death per 100,000 in
the age group f el 1 dr am a t i c a 1 1 y , f r om over 600 annually in
the first h a1f of the 1920-s to 74 at the end of the 1970's,
For the 5-9 age group the decline was from 177 to 34 (Table
2) .
(TABLE 2 ABOUT HERE)
To demonstrate the growing inequality of death rates by
ge nde r , relative rate ratios (RR) were calculated for single
years of age from 1921 to 1979.
To simp1 ify the analysis,
Tab1e 3 present s a portion of the results, for selected
years, and single years of age,
expressed in terms of the
death rates of males over the rates of females.
This figure
indie a t es the degree to wh i ch mi al e mor tali ty i s excess i ve ,
u s i n g the female mortality rate as the "standard" for any
g i v e n age and time period.
(TABLE 3 ABOUT HERE)
Underlying this calculation is the assumption that
biological differences in mortality risks between the sexes
are very small at childhood ages.
The differences which
emerge, then, are mainly the result of differential
socialization patterns which give rise to different
gender-based roles.
A low RR thus indicates that mortality
risks are shared rather equally between the sexes, while a
t h a t social ized elements of "maleness"
high ratio shows that
i mpar t an elevated r i sk of mor tai i ty on young boys.
In
epidemiological terms this is a comparison of two population
"e xpuse d" to the risk of "ma 1eness", and the
groups, one "exposed"
(2) .
other "unexposed"" (2).
In fact, as will be discussed later,
i
s
this distinction is too.finely drawn, and it can be argued
—
that girls also suffer excess mortality by being exposed to
some of the same risks which are only more prevalent among
boys rather than being confined to them exclusively.
The very smal1 numbers of deaths involved, part i cu1 ar 1y
at the late childhood ages, causes the rate ratios to
fluctuate substantially.
Nonetheless the general pattern is
clear.
Relative risks have risen over time, and are higher
at older ages than at young ages during each time period.
The odd anomalies, such as the 9 year-olds in 1975-79, are
caused by the random fluctuations of very small numbers.
The general pattern is for the female death rates to fall
more rapidly than male death rates over time, producing an
increasing RR. The three tables taken together demonstrate
the assertion made at the outset: that the huge decline in
the death rates of young children has been accompanied by a
major rise in the inequality of mortality between the sexes
in those age groups. Why has this occurred?
CAUSES OF DEATH IN CHILDHOOD
The decline in mortality registered over the past 70
years has been brought about largely through the control of
infectious diseases.
Routine vaccinations, antibiotics and
preventative health care measures have made the first few
years of life safe to a degree unparalleled in human
history.
Children are far less likely to become ill,
illness is less likely to be serious, and even serious
illness is more likely to be cured, than was the case sixty,
or even twenty years ago.
The relatively few child deaths
which occur today tend to be caused by external forces, such
as road accidents, drownings, falls, poisonings and homicide
h an iinfectious
n f ec t i ous or non-infectious diseases.
rather tthan
Figure 1
h e cconvergence
on v e r ge n c e o
shows tthe
off the major cause groupings over time
for both males and females combined. Whereas all forms of
disease and biological conditions accounted for over 90
percent of child deaths in the first decades of the century,
today their share has fallen to half, and the proportion of
deaths caused by accidents has risen accordingly.
(FIGURE 1 ABOUT HERE)
Table 4 shows a more detailed breakdown of the numbers
of deaths for major causes of death in 1930 among children
aged 1-9.
Infectious and parasitic diseases cannot be
regarded as a major cause grouping of child deaths today.
However, in 1930 they struck down twice to three times as
many boys as girls.
Superficial 1y this would seem to lend
support to a hypothesis that females are innately stronger
in fending off infections.
However, when the detailed
classifications of infections are examined (Table 5) it i s
-3-
seen that most of the male deaths occurred as a result of
certain bacterial and viral infections, including viral
hepat it i s.
There is no discernable gender bias in other
types of infection.
(TABLE 4 ABOUT HERE)
(TABLE 5 ABOUT HERE)
Malignant neoplasms, congenital anomalies and a number
of non-infectious respiratory diseases stand out among
diseases which continue to be fatal to young children,
By
and large these causes are related to factors endogenous to
the child, rather than parental or child behaviour. They
tend to affect boys more frequently than girls.
Of course,
•=■ ome c on d i t i on s su c h as as t hma c an be aggr av a ted by
environmental factors such as parental smoking or diet.
It
is also clear that many health problems of early childhood
may be related to the behaviour and condition of the motherjust prior to and during pregnancy. To the degree that such
behaviours are common in society, they appear to be related
to a very small number of child deaths.
It is uni
unlikely
i k e 1 >- that
they would affect boys and girls differently.
The broad category of accidents, poisonings and
violence contain fully half the deaths, of children,
accounting for near!?/ ROD deaths in 1980.
Table 6 shows
that of these deaths, most are caused by motor vehicle
accidents, whi1e, especially for the group aged 1-4,
drowning and suffocation remain major killers.
In general
the risk of death from motor vehicle accident rises with
age, while death from drowning or suffocation declines.
Th i>s is because the former group of causes involves an
increasing exposure to risk over time, while the latter
involves declining risk as children become more skilled in
deal i ng with common dangers in their immediate environment,
such as fire, poisons and water.
In both cases, though, the
cruc i al factor is exposure to risk, and as the figures
demonstrate, boys are substantial 1y more likely to die of
these causes, implying that they are more frequently exposed
to the dangers of accident than are girls.
(TABLE 6 ABOUT HERE’)
DISCUSSION
The examination of official statistics of child deaths
in Australia covering the period from 1907 to 1984 reveals
that wh i 1e the overal 1 death rates have dec 1 i ned
dr amati cal 1y, the sex d i fferen t i als in child mor tai i ty have
been increasing. These are in part related to biological
differences between males and females which imply that males
-4-
are at a somewhat greater risk of death from congenital
abnormal i ties, diseases of the respiratory system and
congenital abnormalities,
The major factor underlying this
grow i n g differential appears to be gender specific
behavi oural factors.
In socializing children parents and
1 a t e r teachers treat boys and girls very differently.
Th i s
conditions their exposure to injury and death.
Among the
factors which appear salient:
1. Parents and teachers discourage aggress i ve be h av i our
in girls while accepting or encouraging it in boys. (3)
2. Parents tend to watch over and protect girls more
systematical 1y than boys, perhaps out of a sub-concious fear
that girls are more vulnerable to abduction or violation.
3. In late childhood, boys are more likely to have
bikes,
ride bikes on the road, and travel long distances
away from home, alone or in groups, on foot or by bike (4).
4. Risk taking behaviours learned in early childhood
are reinforced through patterns of peer pressure in later
childhood which magnify gender differences C5).
These factors all imply that boys are differential 1y
exposed to risks which can potentially result in injury or
death. This exposure is Jointly determined by less parental
and teacher control, and more aggressive and dangerous
behavi our among boys.
Mortality rates, though historically low, r e v e a 1 a
•=. i gn i f i cance .
problem of great social significance.
Death is but the rare
conclusion of a common process, that of socializing boys and
girls differently with regard to risk-taking and aggressive
behaviour.
The mortality differentials at this age group
are thus significant in two other dimensions.
First^ t h e y
imply that programmes for controlling morbidity and
mortality in childhood should concentrate on modifying the
gender role formation process in homes and in schools at
very young ages, with particular emphasis on breaking the
1 ink between the notion of maleness and aggressive and risky
behaviourSecond, researchers into sex differentials in
adolescent and adult mortality should direct more attention
to the ways i n wh i ch th i s mor tai i ty i s cond i t i oned by
behaviours developed in childhood.
This might have
significant impl icat ions for a variety of campaigns designed
to promote better health through behavioural modification.
Ac k n qi.--! 1 e d q e m e n t •=•
-5-
R i chard
c a 1 c u 1 a t i on s
time to t i me
up necessary
here .
Choo carried out a long series of laborious
to produce the tables for this paper.
From
he was helped by Pat Quiggin, who also chased
references, not all of which could be cited
References
1 . Office of the Australian Government Actuary. Au str-al i an
Life Tables: 1980-82. Canberra: Australian Government
Fu b1 i sh i n g Service, 1 985.
2. Last, John M. (ed.) A Dictionary of Epidemiology , Oxford:
Oxford University Press, 1983.
3. R i v ar a, Frederick P. "Epidemiology of Childhood Injuries:
II. Sex Differences in Injury Rate s . " Amer i c an J ou r- n a 1 of
Diseases of Children. 136<6):505. 1982.
4. Rivara, Frederick P. "Epidemiology of Childhood Injuries:
I. Review of Current Research and Presentation of
Conceptual Framework." American Journal of Diseases of
Ch i1dren, 136(5):401. 1982.
5. Levjis, Charles E and Mary Ann Lewis. "Peer Pressure and
Risk-Taking Behaviors in Children." Am e r i c a n J c» u r n a 1 o f
Public Health. 74(6):5S3. 1984.
-6-
Tab’e 1 . Sex Ratios o-f Child Deaths in Austral ia , 1907-34,
(Male deaths/Female deaths x
Ye ar s
1907-0?
1910-14
1915-19
1920-24
1925-29
1930-34
1935-3?
1940-44
1945-4?
1950-54
1955-59
1960-64
1965-69
1970-74
1975-79
1930-34
Age
100)
1-4
Age 5-9
110.7
113.2
114.6
120.5
113.7
1 20.8
116.2
124.1
1 28.0
129,5
129.2
119.7
1 26.5
1 34.8
135,7
1 33.9
112.7
111.5
115.2
117.9
123.6
1 36.5
1 38.1
1 38,6
147.8
145.5
151.6
144.3
1 37.6
144.5
141.5
1 80.3
Sources: Various publications of the Australian Bureau of
Statistics and the Australian Data Bank maintained
by the Demography Department, Australian National
University.
_7-
Table 2. Child Death Rates in Australia,
1920-79.
(Deaths per 100,000 mid-period population of the age group)
Age 1-4
Age 5-9
Year s
Boys
Girl s
Boys
G iris
1920-24
1925-29
1930-34
1935-39
1940-44
1945-49
1950-54
1955-59
1960-64
1965-69
1970-74
1975-79
603.7
513.2
421.7
372.9
337.4
203.2
177.9
143.7
109.8
100.1
97.3
74.1
50 6.9
446.9
364.0
332.5
281 .9
165.3
1 43.8
116.5
96.2
83.3
76.0
56.9
177.1
169.1
153.9
150.5
1 38.6
94.0
75.9
57.7
52.2
45.3
44.8
34.2
1 50.3
1 35.0
116.5
113.5
103.4
65.8
54.4
39.8
37.9
34.6
32.6
25.3
Note :
Death Rates for 1907-20 and 1980-84 are not presented
because of incompleteness of population data
related to these periods.
Source: See Table 1 .
-8-
Table 3. Risk Ratios (RR) for Child Death Rates, for
Selected Time Periods, Austral ia.
(RR = Male Death Rate / Female Death Rate)
Age
1925-29
1955-59
1975-79
1
2
3
4
5
6
7
8
9
1.15
1.15
1 .09
1.19
1.10
1 .25
1.22
1 .46
1.33
1 .22
1 .22
1 .30
1 .21
1.27
1 .49
1.36
1 .69
1.53
1.24
1 .20
1.50
1 .36
1 .50
1.36
1 .39
1 .39
1.19
Source: See Table 1 .
-9-
Table 4. Major Causes of Child Deaths in Australi a, 1980
Age Group:
Sex :
1-4
Mai es Females
All Deaths:
Major Cause Croups:
infectious and Parasitic
Malignant Neoplasms
Endocrine, nutritional and
metabolic diseases and disorders
Disease of blood and blood
form inq organs
Diseases of nervous system and
sensory organs
Diseases of the circulatory
system
Diseases of the respiratory
system
Diseases of the digestive
system
Diseases of the genitourinary
or skin or skeletal system
Congenital abnormalities
Conditions originating in the
per i natal per i od
Symptoms, signs and ill-defined
condit i ons
Accidents, poisonings and
violence
5-9
Males Females
308
215
211
144
12
32
6
25
12
39
4
40
6
1
3
8
1
0
1
2
18
18
10
8
7
7
5
4
23
16
7
3
4
4
2
1
44
1
20
1
15
0
11
4
0
0
0
6
6
3
1
150
111
113
61
10
I ab1e 5. Child Deaths Caused by Infectious and Parasitic
Disease: Detailed List of Causes, 1980.
Age Group
Sex:
1-4
Males Females
5-9
Hal es Ferna1es
Group Iota!:
Detailed Causes!
12
6
12
4
Intestinal
I ubercu1osi s
Bac teria 1
Po1i owyeli t i s
Viral diseases
Other infectious and parasitic
2
3
0
1
2
0
0
1
0
0
0
i
1
1
1
1
3
0
6
0
11
3
5
1
I a b 11? 6.
Child Deaths Caused by Accidents Poisonings and
Violence; Detailed List of Causes, 1980.
Age Group:
Sex:
1-4
Mai es Females
5-9
Hales F 8<fta 1 es
Group Totals
Detailed Causes:
150
111
113
61
Railway accidents
Water transport accidents
Accidental poisonings
Medical misadventure
Accidental falls
Accidents by fire and flames
Natural and Environmental
Drowning, suffocation
Other accidents
Drug reactions
Homicide and other purposely
inflicted
Motor vehicle accidents
Other road vehicle accidents
2
0
4
0
3
6
5
66
9
0
0
0
4
1
4
4
0
41
9
1
1
5
0
1
2
2
3
14
9
0
0
1
1
0
1
0
1
4
5
0
8
47
0
4
43
0
4
72
0
1
45
2
0
1
0
0
1
25
9
4
4
12
1
17
5
3
1
11
15
29
5
5
4
12
4
2/
1
1
0
Of Motor Vehicle accidents:
Collision with train
Collision with another motor
vehi c1e
Collision with other vehicl e
Collision with pedestrian
Collision on highway
Loss of control on highway
Other motor vehicle accidents
12
FIGURE 1 : The Relative Role of Diseases and Biological
Conditions compared to Accidents as Causes of
Child Deaths in Australia (1911 - 1980)
AGE 1 - 4
100-•
90
Diseases and Biological Conditions
80-70-w
X
H
60-’
<
M
Q
50-’
►J
<
40--
Ph
o
w
e
<
30-.
2;
w
o
w
pp
20--
Accidents
10— i—
1911
4----------- ♦—
♦
1966
1947
1980
AGE 5-9
100 -
90-80tn
K
H
3Q
fJ
fJ
70-
Diseases and Biological Conditions
60
50-
PH
o
PJ
o
<
H
Z
M
O
Pd
PJ
Ph
40Accidents
30-
20
10
+
1911
1947
1966
1980
Research Note on
CHILD
SURVIVAL
Number
14CS
Date
23 July 1987
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
THE 1985 INTERCENSAL SURVEY OF INDONESIA:
4. INFANT AND CHILD MORTALITY LEVELS
Kim Streatfield
and
Ann Larson
International Population Dynamics Program
Department of Demography
The Australian National University
Note:
- \
• 3".. x .
?
W
-*?
-
ANO
Child Survival Research Notes are brief discussions of
issues of current relevance to researchers and policy
makers concerned with problems of high infant and child
mortality in the world. The International Population
Dynamics Program, Department of Demography, The Australian
National University, distributes these notes with their
regular Bibliographic Circular. Production of the Child
Survival Research Notes is made possible through a grant
from the Ford Foundation (840-0893). Responsibility for
the content of Child Survival Research Notes rests with
the author(s) alone, and not the above-listed organisations.
Research Note on
CHILD
SURVIVAL
Number
14CS
Date
23 July 1987
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
THE 1985 INTERCENSAL SURVEY OF INDONESIA:
4. INFANT AND CHILD MORTALITY LEVELS
Kim Streatfield
and
Ann Larson
International Population Dynamics Program
Department of Demography
The Australian National University
Note:
Child Survival Research Notes are brief discussions of
issues of current relevance to researchers and policy
makers concerned with problems of high infant and child
mortality in the world. The International Population
Dynamics Program, Department of Demography, The Australian
National University, distributes these notes with their
regular Bibliographic Circular. Production of the Child
Survival Research Notes is made possible through a grant
from the Ford Foundation (840-0893). Responsibility for
the content of Child Survival Research Notes rests with
the author(s) alone, and not the above-listed organisations.
INFANT AND CHILD MORTALITY LEVELS
INDONESIA, 1985
Since the early 1970’s Indonesia has implemented a
number of health intervention and related programs which
might be expected to have brought about a reduction in
infant and child mortality.
The widespread uptake of family planning alone, with
its consequent decline in the numbers of high risk births
(high birth order, high maternal age), would certainly
have reduced infant mortality, even without any other
changes in health conditions.
The UNICEF GOBI-FFF
Program being implemented through the Ministry of Health
has greatly increased access to growth monitoring
facilities, to immunization, to oral rehydration therapy
for diarrhoeal diseases, to education about nutrition and
breastfeeding, and to food supplements for non-thriving
children.
These programs have been developing in an
environment of increasing levels of formal education,
rapidly spreading communication networks, and. a general
increase in levels of economic development.
The first opportunity to estimate national levels of
infant mortality in Indonesia came from the 1971
Population Census. The use of the Brass indirect method
with data on numbers of children ever born and children
still living for each woman produced an estimated infant
mortality rate for Indonesia (urban + rural) of 141 per
1,000 live births for the period 1960-70 (Cho et al.,
1980:20). A re-estimate of the rate by Hull (1978:10)
using the Trussell indirect method, gave a fairly
consistent figure of 143 per 1,000 for the year 1968.
The indirect method mentioned above has been very
important in providing a picture of fertility and infant
mortality levels in the large parts of the developing
world where vital registration of births and deaths is
inadequate or non-existent. The method depends, however,
on meeting certain assumptions of constant fertility and
mortality, of stable patterns of marriage, and of
relatively uncontrolled high fertility.
While there
have been claims that the method is not particularly
sensitive to changing mortality, there is now some
evidence to suggest that in conditions of rapid mortality
and fertility decline, the indirect methods may
considerably overestimate infant mortality (Sullivan and
Wilson, 1982:83).
It is in these circumstances of uncertain stability
or robustness of the indirect methods of estimating
infant mortality that another method has been developed.
With the ’last birth’ method, mothers are asked about the
date of birth of their last born child, then about the
survival status of that child. Estimates of proportions
of children surviving to the survey or census date can
1
then be made, and by comparison with model life tables,
infant mortality rates can be estimated.
This method is basically the same as that described
by Hull and Dasvarma (1987) for estimating fertility
levels, and is subject to the same kinds of reporting
errors. These potential errors are: ”(1) the possibility
that mothers who had a birth in 1985 might have
erroneously dated the birth as having occurred in 1984,
(2) failure by the interviewer to ask the question or
accurately record the reply, (3) failure of the mother to
report the date of the last birth through forgetfullness
or desire to evade the question." (ibid:2).
Hull and
Dasvarma go on to state that these types of potential
errors were not thought to have affected the quality of
the SUPAS 1985 data.
It is a little difficult to be
certain of this because the Central Bureau of Statistics
has apparently redistributed the ’Not Stated’ responses
on date of, and birth order of last birth before
publishing the figures.
This was not done with the 1980
Census data.
The key difference between fertility and mortality
estimation relates, of course, to the reporting of deaths
of last born children.
It is suspected that if a child
born within the last year died very young and was
followed soon after by the birth of a surviving child,
then the birth and death of the penultimate child would
very likely go unreported.
This failure to mention the
death of that child may be unintentional or intentional
if, as in some cultures, it is considered inauspicious to
discuss a dead child. Hull and Dasvarma argue that this
probably does not occur to any great extent in Indonesia,
or at least in Java due to the necessity to conduct
public ceremonies following a death (1987:3).
The Intercensal Survey of 1976 (SUPAS) provided the
first opportunity in Indonesia to use the ’Last Birth’
method developed by Hull (1978). This method produced
infant mortality rates for Indonesia of 107 per 1,000 for
1975, a figure which accords well with the result of 112
per 1,000 from the Trussell method (Hull, 1978:10).
When adjustment was made for the possibility of a mother
having had a penultimate birth in the reference period of
1.2 years (January 1, 1975 to March 15, 1976) which
subsequently died young and was followed by another birth
prior to the survey, the infant mortality rate was 114
per 1,000. The possible reasons for this relatively high
figure was discussed by Hull (1978:11).
The 1976 SUPAS
survey was not designed to produce estimates for
individual provinces outside Bali, and the most populous
island, Java.
The 1980 Population Census with its five percent
sample of the total population was large enough to be
used to provide estimates for individual provinces.
Estimates were made using both ’last birth’ method and
the Trussell indirect approach (Table 2) (see Soemantri,
2
1983:188 for comparison with 1971 levels),
The infant
mortality rate for all Indonesia was 107 per 1,000
according to the Trussell approach using West model life
table.
The 'last birth’ method gave an estimate of 97,
per 1,000. The estimates using the ’last birth method
are expected to usually be lower than the indirect
technique estimates.
In Table 2, some 20 of the
provinces are in that expected direction.
RESULTS
The data from the second Intercensal Survey, SUPAS
The
1985 (Table 1), suggest a major decline since 1980.
Trussell indirect method gives an estimate of 72 per
1,000 for all Indonesia, some 33 percent lower than the
1980 indirect estimate. The figures in Table 1 show the
infant and child mortality estimates using both CoaleDemeny Model Life Tables and the more recent United
Nations Model Life Tables. There is not a great
variation among the different patterns tested, and there
is close agreement between the two patterns most often
used for Indonesia, the Coale and Demeny West model, and
the U.N. General model. These two give an infant
mortality rate of 72 per 1,000 for women aged 25-29,
probably the preferable age range to be used when numbers
in the 20-24 year age range are limited and estimates may
be unstable, as in a sample survey. This is despite the
earlier reference period for the 25-29 year age range
estimate.
The instability in the 20-24 age estimates is
reflected in the estimates where IMRs from the 20-24 age
range are considerably higher than those from the 25-29
age range in urban areas, and slightly higher in rural
areas.
Somewhat surprisingly, the ’last birth’ method gives
an estimate of 72.2 per 1,000 for 1985. This figure
matches closely the indirect estimates for considerably
earlier periods, suggesting either plateauing of infant
mortality during the 1980’s, or an error in one or both
estimates. The ’last birth’ estimates are a little less
stable than the Census estimates, as might be expected
with a smaller sample. There are 14 provinces with
higher rates than the indirect estimates for an earlier
period.
Although the correlation between the indirect
and the ’last birth* estimates is lower in 1985 (r=0.77),
than in 1980 (r=0.83).
The pattern of change in provincial infant mortality
rates also depends greatly on which method of estimation
is used.
The Trussell method produces a range of
provincial annual percentage declines over the five year
period of 5.0 percent in West Nusatenggara to 22.2
percent in Irian Jaya.
These figures compare with the
’last birth’ estimates ranging from stability in Central
Java (unlikely to be true), a slight increase (0.8%) in
Maluku to a decline of 15.5 percent in East Kalimantan.
This reflects the much wider range of change with the
’last birth' estimates than for the indirect estimates.
3
While the uncertainty attached to the ’last birth’
estimates reflected in the apparently erratic patterns of
the declines, the overall impression of the declines is
that the island of Sumatra has experienced the greatest
declines, with the rates in four of its eight provinces
falling by more than 10 percent per year. The island of
Sulawesi also appears to have undergone a substantial
decline according to the indirect estimates, but this
pattern is not supported by the ’last birth’ estimates.
The final table (Table 3) shows provincial child
mortality rates using the West model Coale and Demeny
Tables and the U.N. General Model Tables. The MortPak
computer program used to generate these estimates
actually produces 4ql values, or probabilities of
survival from exact age one to exact age five years,
rather than actual central death rates.
The data show a
range from about 9 or 10 per 1,000 in Yogyakarta (and
Irian Jaya!) to a high of 85 to 98 per 1,000 in West
Nusatenggara (varies with Model Tables used).
The
average for all Indonesia was about 30 per 1,000 children
aged one year in late 1981 (Table 1).
If the infant and child mortality probabilities are
correct then it suggests that the age pattern of
mortality at young age may be changing with greater
impact on child survival being experienced by infants
rather than children.
This may be consistent with
intervention programs targetting pregnant women with
tetanus toxoid immunization, infants with immunization
against various serious diseases, prolonged
breastfeeding, etc., but not yet targetting the serious
respiratory and parasitic diseases suffered by toddlers
and young children.
In conclusion, the data indicate that infant
mortality has been halved since the late 1960’s.
This
is a major achievement, and as a Government target for
the end of the fourth Five Year Development Plan (ending
March 1989) (BAPPENAS, 1984:54), and suggests that
further improvements in child survival may reasonably be
hoped for.
4
REFERENCES
BAPPENAS (National Development Planning Agency) (1984)
’Policies and Prospects for Sustained Development under
Challenging Conditions’, The Fourth Five Year Development
plan of Indonesia, 1984/85-1988/89.
Cho,L-J, Suharto,S., McNicoll.G., and S.G.M. Mamas
(1980), PopulationGrowthofIndonesia, Monograph of the
Center for Southeast Asian Studies No.15, Kyoto
University, University Press of Hawaii, Honolulu.
Hull,T. (1978) 'An Estimate of Infant Mortality in
Indonesia in 1975', Working Paper Series No.10,
Population Studies Center, Gadjah Mada University,
Yogyakarta, Indonesia.
Hull,T. and G.L. Dasvarma (1987) ’The Intercensal Survey
of Indonesia: 3. Evidence of Continuing Fertility
Decline’, Research Note No.77, 9 July, pp.7. .
Soemantri,S. ’Trends and Regional Differentials in Infant
Mortality Rates’, in Seminar Tingkat Kematian Bayi di
Indonesia, 1-3 Februari, 1983: 173-192.
Sullivan,J.M. and S.E.Wilson (1982) ’The 1980 Baseline
Round of the East Java Population Survey: A Final
Report', Technical Report No.1, Inti. Program of
Laboratories for Population Statistics (POPLAB), The
University of North Carolina, Chapel Hill, NC, and
Central Bureau of Statistics, Jakarta.
5
Table 1: Infant and Child Mortality Rates (Per 1,000) for
Indonesia (Urban & Rural; Urban; Rural), Coale and Demeny
Models, and United Nations Models, SUPAS 1985.
Reference
date
COALE-DEMENY MODELS (Trussell equations)
South
East
North
West.
INDONESIA
Infant Mortality
70
72
63
68
20-24
Oct 1983
75
78
64
25-29
Nov 1981
72
Child Mortality
22
19
37
20-24
Oct 1983
28
26
22
38
30
25-29
Nov 1981
Age of
woman
URBAN INDONESIA
Infant Mortality
20-24
Dec 1983
25-29
Mar 1982
Child Mortality
20-24
Dec 1983
25-29
Mar 1982
RURAL INDONESIA
Infant Mortality
20-24
Sept 1983
25-29
Oct 1981
Child Mortality
20-24
Sept 1983
25-29
Oct 1981
65
49
61
44
69
53
67
52
26
16
35
21
17
10
20
12
78
74
72
66
82
81
79
77
34
32
44
39
24
23
28
27
UNITED NATIONS MODEL (Palloni-Heligman equations)
South Far
Asian
East
Lat Am Chilean
INDONESIA
Infant Mortality
67
67
74
66
65
20-24
Sept 1983
72
71
72
82
70
25-29
Feb 1982
Child Mortality
27
28
29
15
31
20-24
Sept 1983
31
30
32
18
35
25-29
Feb 1982
URBAN INDONESIA
Infant Mortality
20-24
Nov 1983
25-29
May 1982
Child Mortality
20-24
Nov 1983
25-29
May 1982
RURAL INDONESIA
Infant Mortality
20-24
Sept 1983
25-29
Jan 1982
Child Mortality
20-24
Sept 1983
25-29
Jan 1982
63
48
70
55
64
49
63
49
64
49
29
19
13
9
27
17
25
16
26
16
74
73
85
86
76
74
76
75
76
74
39
37
19
19
36
34
33
32
35
33
6
Table 2: Provincial Infant Mortality Rates (Per 1,000)
Estimated from SUPAS 1985 using Last Birth Method and
Trussell Indirect Method (West Model Life Tables).
-Infant Mortality Rate-- Annual Percentage
-- 1980-----1985-Decline 1980-85
L.B.
Tr.
Lt.B.,... Tr
___ _ Tr.5.
Indonesia
97
107
72.2
72
-5.9
-7.9
Sumatra
D.I. Aceh
North Sum
West Sum
Riau
Jambi
South Sum
Bengkulu
Lampung
121.7
71.8
111 .4
112.2
120.7
95.1
96.1
69.9
91
89
121
113
118
98
106
97
63.0
65.5
56.1
54.1
87.2
80.9
57.8
49.7
45
58
77
59
62
71
62
59
-13.2
-1.8
-13.7
-14.6
-6.5
-3.2
-10.2
-6.8
-14.1
-8.6
-9.0
-13.0
-12.9
-6.5
-10.7
-9.9
Java
54.5
Jakarta
107.3
West Java
Central Java 65.4
Yogyakarta
43.3
91.6
East Java
80
129
96
62
99
26.4
82.0
68.7
<23
80.4
32
84
70
37
71
-14.5
-5.4
+ 0.01
-2.6
-18.3
-8.6
-6.3
-10.3
-6.7
74.2
169.9
105.2
88
187
124
52.6
142.0
61.5
79.7
64
146
88
73
-6.9
-3.6
-10.7
-6.4
-5.0
-6.9
Kalimantan
West Kai
Central Kai
South Kai
East Kai
107.7
108.2
134.9
100.9
116
100
121
99
89.5
87.5
64.4
46.6
54
73
88
42
-3.7
-4.3
-14.8
-15.5
-15.3
-6.3
-6.4
-17.2
Sulawesi
North Sul
Central Sul
South Sul
S.E. Sul
79.7
136.7
95.9
106.3
94
128
108
114
55.3
110.9
70.9
99.8
55
78
65
78
-7.3
-4.2
-6.0
-1.3
-10.7
-9.9
-10.2
-7.6
Maluku
Irian Jaya
104.6
130.8
124
106
109.1
30.1
80
35
+ 0.8
-29.4
-8.8
-22.2
Nusatenggara
Bali
West NT
East NT
East Timor
NB:
L.B.: estimate is based on ’Last Birth’ method.
Tr. : estimate is based on indirect method of
Trussell (West MLT).
7
Table 3 — Provincial Infant and Child Mortality Estimates (Proportions
Expected to Survive).
Coale - Demeny West Model
U.N.
General Model
Province
Reference
date
Infant
Mor
tality
Child
Mor
tality
Reference
date
Infant
Mor
tality
Child
Mor
tality
Sumatera
Aceh
Sumatera Utara
Sumatera Barat
Riau
Jambi
Sumatera Selatan
Bengkulu
Lampung
Jan 1982
Apr 1982
May 1982
Mar 1982
Oct 1981
Apr 1982
Feb 1982
Dec 1981
0.045
0.058
0.077
0.059
0.062
0.071
0.062
0.059
0.013
0.021
0.030
0.022
0.024
0.029
0.024
0.020
Apr 1982
July 1982
July 1982
May 1982
Jan 1982
June 1982
May 1982
May 1982
0.045
0.057
0.077
0.059
0.062
0.071
0.062
0.059
0.014
0.021
0.035
0.022
0.024
0.030
0.025
0.022
Jawa and Madura
DKI Jakarta
Jawa Barat
Jawa Tengah
DI Yogyakarta
Jawa Timur
Nov 1981
Aug 1981
Dec 1981
June 1982
Sep 1981
0.032
0.084
0.070
0.037
0.071
0.007
0.038
0.029
0.009
0.029
Feb
Nov
Mar
Aug
Dec
1982
1981
1982
1982
1981
0.031
0.083
0.070
0.036
0.071
0.008
0.040
0.030
0.010
0.030
Nusa Tenggara
Bali
Nusa Tenggara Barat
Nusa Tenggara Timur
Timor Timur
Mar 1982
Nov 1981
June 1982
July 1981
0.064
0.146
0.088
0.073
0.025
0.085
0.041
0.031
June 1982
Feb 1982
July 1982
Dec 1981
0.064
0.141
0.087
0.074
0.026
0.098
0.043
0.033
Kalimantan
Kalimantan Barat
Kalimantan Tengah
Kalimantan Selatan
Kalimantan Timur
Nov
Nov
Oct
Oct
1981
1981
1981
1981
0.054
0.073
0.088
0.042
0.018
0.030
0.041
0.011
Feb
Feb
Jan
Jan
1982
1982
1982
1982
0.053
0.073
0.088
0.041
0.019
0.032
0.044
0.012
Sulawesi
Sulawesi
Sulawesi
Sulawesi
Sulawesi
Jan
Dec
Mar
Nov
1982
1981
1982
1981
0.055
0.078
0.065
0.078
0.020
0.034
0.026
0.034
Apr
Apr
May
Feb
1982
1982
1982
1982
0.055
0.078
0.065
0.078
0.020
0.036
0.026
0.036
Maluku
Nov 1981
0.080
0.035
Feb 1982
0.080
0.038
Irian Jaya
Apr 1981
0.035
0.009
Sep 1981
0.035
0.010
Utara
Tengah
Selatan
Tenggara
8
Research Note on
Number ijcs
Date
CHILD
SURVIVAL
26 November 1987
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
TRENDS IN DRINKING WATER AND SANITATION FACILITIES
INDONESIA 1980-85, AND IMPACT ON INFANT AND
CHILD MORTALITY IN EAST JAVA
Kim Streatfield
and
Miranda Korzy
Child Survival Project,
International Population Dynamics Program,
Department of Demography,
The Australian National University
Note:
LIBRARY
■J (
AND
INFORM* TON
C ENTf f
A >
Child Survival Research Notes are brief discussions of
issues of current relevance to researchers and policy
makers concerned with problems of high infant and child
mortality in the world. The International Population
Dynamics Program, Department of Demography, The Australian
National University, distributes these notes with their
regular Bibliographic Circular. Production of the
Child Survival Research Notes is made possible through
a grant from the Ford Foundation (840-0893).
Responsibility for the content of Child Survival
Research Notes rests with the author(s) alone, and not
the above-listed organisations.
1. BACKGROUND
The Indonesian Government subscribes to the general
philosophy of WHO’s ’’Health for All by the Year 2000”. As
part of this approach the Government is expanding supplies of
clean drinking water and higher quality sanitation facilities
across the country. The target for the fourth Five-Year
Development Plan (REPELITA IV 1984/85-88/89) is to expand
supplies of potable (drinkable) water from 1983 levels of 32%
of rural households and 60% of urban households, to levels in
1988 of 55% (rural) and 75% (urban) households (BAPPENAS,
1984:54). No specific targets were mentioned for sanitation
facilities, though broad improvement in community hygiene was
mentioned as a target (ibid).
Data on these very important facilities have now been
collected in the 1980 Population Census and the 1985
Intercensal Survey (SUPAS 85). Thus it is possible to begin
to
examine
the ~impact
of this1 4-kexpansion
on health,
i
_______ _ — —
~ X*
tn
i r-\tntm^tnxr\
basic- assumption
of many health
care
(primary) programs that
the provision of clean drinking water and sanitation will lead
to a decline in infant and child mortality. The effect may be
more marked on child than infant mortality if infants are
primarily breastfed, as they are less likely to be exposed to
diarrhoea-causing agents under such circumstances.
j
In this paper the infant mortality and child mortality
rates according to source of drinking water, and type of
sanitation facility, will be presented based on data from the
1980 Census. Without access to the data (raw) computer tapes
from SUPAS 1985 it has not yet been possible to calculate the
1985 rates - however it is possible to obtain the 1985
distributions of drinking water and sanitation facilities.
While there has been a substantial decline in infant and
child mortality in Indonesia in the 5 year reference period
(IMR fell from 97 to 72 per 1,000 live births, see Streatfield
& Larson, 1987), there have been a variety of changes which
have probably contributed to those declines. These changes
include expanded health services providing child immunization
and other aspects of primary health care, improved nutrition,
smaller proportion of births in the higher risk categories,
etc.
Here the 1980 infant and child mortality rates will be
applied to the 1985 distribution of categories of water supply
and sanitation facilities to hypothesize what the overall
provincial infant and child mortality rates would have been in
1985 if no changes had taken place other than in the
distribution of such facilities. Alternatively, the results
could be interpreted as what would the overall infant
mortality and child mortality rates have been in 1980 if
Indonesia had the 1985 distribution of water supply and
sanitation facilities, in other words, what effect would
expanding these facilities to 1985 levels have had on
1
mortality levels if the expansion occurred instantaneously in
1980.
This is one simple approach to trying to estimate the
relative contributions of different changes and health
interventions to the mortality decline during/over the period.
It is analogous to decomposing fertility change into
proportional contributions of changes in age structure, in
marriage patterns and in marital fertility.
Ideally the other
contributors to the 1980-85 mortality decline should also be
tested in the same way, and work is under way on the effect of
the fertility decline.
2. METHODOLOGY
The data set from the 1980 Indonesian Census is very
large, thus we have chosen to work with only one province,
East Java, for this study. With a population of 29.2 million
in 1980, East Java is larger than many countries and is
considered sufficiently large to provide reasonable estimates
of mortality rates according to the category of health
facility under study here.
The infant (IqO) and child mortality (4ml) rates have
been calculated using the UN MORTPAK-LITE 2.0/NCP program for
micro-computers*.
The rates used here are taken from the maternal age group
20-25 years and the West Coale-Demeny model is used. While
there has long been varied opinions on whether West is the
most appropriate model for Indonesia, the results here would
be little affected by choosing South (C&D), or the UN General
Pattern, as we are concerned primarily with relative levels of
mortality.
3. RESULTS
The changes in distribution of drinking water and
sanitation facilities in East Java are presented in Table 1.
The period 1980 to 1985 has seen a substantial growth in
the number of households, 13.37% or 2.5% p.a., compared to a
1.4% p.a. growth in population. In terms of facilities, the
866,046 household increase has involved considerable
increases in higher quality facilities of both kinds - in
piped, bore, and well sources of drinking water, and in
private sanitation with septic tank. The absolute decreases
in numbers of households relying on river water and rainwater
* This very useful program was prepared by the United Nations
Population Division of the Department of International
Economic and Social Affairs and can operate on IBM-compatible
micro-computers using 2 floppy disk drives or single floppy
and hard disk. Contact persons for MORTPAK-LITE are Larry
Heligman, John Kanakos and Sharon Kirmeyer at United Nations,
New York, N.Y. 10017.
2
is an indication of the expansion of potable water services to
such households by the Government. The very impressive
increase of 374,598 households having access to piped water
reflects a major effort to achieve the above mentioned
targets.
Table 1.
Distribution (%) of Drinking Water Facilities and
Sanitation Facilities, East Java, 1980-1985 (U+R)
Percentage Distribution
1980____________
1985
Drinking Water Facility
8.08
Pipe/tap
2.26
Pump/bore
67.08
Well
15.74
Spring
4.53
River
0.37
Rain
1.83
Other
0.10
Not Stated
100%
6,478,680
12.23
4.53
63.08
14.95
3.17
0.05
1.99
0.01
100%
7,344,726
Sanitation Facility
Private Toilet with
7.89
Septic Tank:
Private Toilet without
21.68
Septic Tank:
Shared/Public/
70.09
Other:
0.00
Not Stated:
100%
6,478,680
Absolute
Change
1980-85
+374,598
+185,681
+286,655
+78,385
-60,522
-20,472
+27,298
-5,577
+866,046
(=13.37%)
12.99
+442,998
22.72
+264,057
64.29
0.00
100%
7,344,726
+180,901
-21,910
+866,046
It is difficult to be sure exactly what the Government
defines as ’’potable water” as the REPELITA figures do not
match exactly the total Indonesia figures for total proportion
of people with access to piped water or pump water, or well
water (Table 4). Pipe and pump water is probably assumed to
be safe to drink, but water taken from wells may or may not be
safe to drink without boiling, depending on whether or not the
well is covered, and how close the well is to sanitation
facilities and other potential sources of contamination.
The increase of 442,998 households with a private toilet
with septic tank again reflects a major effort over five
years, althought the 1985 proportion of households (13%) with
such high quality facilities is still rather low, there are
still almost 5 million households in East Java without any
private toilet facilities.
3
3.1 Mortality Rates According to Drinking Water and
SanitationFacilities
The infant and child mortality rates associated with
categories of drinking water and sanitation facilities, are
presented in Table 2.
Table 2.
Infant Mortality and Child Mortality Rates According
to Category of Drinking Water Source, and Sanitation
Facility, East Java, 1980.
Inf ant
Mortality Rate
Drinking Water Facility
Pipe/tap
Pump/bore
Well
Spring
River
Rain
Other
Sanitation Facility
Private Toilet with
Septic Tank:
Private Toilet without
Septic Tank:
Shared/Public/Other:
Note:
Child
Mortality Rate
97
103
102
115
130
137
96
47
51
51
60
72
77
46
80
36
95
111
43
58
Infant Mortality Rate is deaths per 1,000 live births,
Child Mortality rate is deaths per 1,000 children aged
1-4 years.
Drinking Water.
The patterns are quite consistent with the usual
assumptions that piped water is the least likely to be
contaminated with bacteria, etc. The identical rates
associated with water from a bore or pump (hand operated or
motor-driven), and from a well are slightly surprising,
although it is common practice in Java to boil water before
drinking. This may apply less to those who do not have access
to even a well but rather obtain drinking water from a spring,
river, or from rain water. Economic factors may mean that
these latter households are unable to boil their drinking
water, at some or all times.
It is also surprising how
similar the mortality rates are for three higher quality
sources of drinking water.
It is unexpected that rain water is associated with such
high mortality rates, though it cannot be separated from the
likelihood that those families who rely only on rainwater may
be too poor to afford a more reliable supply such as a well.
Thus economic deprivation, with all its implications, may play
a role here. There is some anecdotal evidence from Yogyakarta
4
that some people find that rainwater lacks ’’taste” and they
have been known to add taste by putting earth into the
rainwater (Hagul, personal communication).
This, and similar
practices, may be widespread, as may be problems with keeping
rainwater stored hygienically through dry periods.
The category ’other’ has a relatively low rate. This
------ probably includes those households which obtain their
group
drinking water from mobile sellers who distribute piped water
in tins or drums, sold by the litre.
Sanitation.
The pattern of IMR associated with sanitation facilities
also fits the expected pattern, ranging from a low of 80 for
households with private toilet with septic tank, to 111 for
’other’ meaning ’in the river’, in the bush, yard, etc.
The category patterns of child mortality rate must, by
definition, follow those of the IMR due to the nature of the
indirect techniques.
3.2 Impact of. Expansion of Prinking Waterand Sanitation
Faci1ities on Infant and Child Mortality
Table 3:
Overall Infant and Child Mortality Rates According
to 1980 and 1985 Distributions of Drinking Water and
Sanitation Facilities
1980
1985
Diff.
% Decline
Drinking Water Facility
Infant Mortality Rate:
Child Mortality Rate.-
105.0
53.1
104.2
52.5
0.8
0.6
0.76%
1.15%
Sanitation Facility
Infant Mortality Rate:
Child Mortality Rate:
104.6
52.8
103.3
51.7
1.3
1.1
1.24%
2.08%
Note:
Infant and Child Mortality Rates as defined in Table 2.
As described above, the figures for infant and child
mortality for the year 1985 are estimated by applying the 1980
category-specific mortality rates for each facility to the
1985 distribution of households across these categories.
The estimates indicate that if the 1985 distribution of
sources of drinking water had existed in 1980, the impact on
infant mortality and child mortality would have been declines
of 0.76% and 1.15% respectively.
If the 1985 distributions of
sanitation facilities had existed in 1980, the infant and
child mortality rates would have been reduced by 1.24% and
2.08% respectively, declines slightly greater than for
drinking water.
5
The reasons for these very small declines are due partly
to the fact that while better quality sources of drinking
water (pipe, pump, well) were increased by 846,934 during the
period (Table 1), most of this enormous effort was absorbed by
the 866,046 increase in numbers of households. Similarly for
the increase of 707,055 in better quality sanitation (private
toilets with or without septic tank).
The other reason is that the infant (and child) mortality
rates according to source of drinking water are relatively
similar for the assumed higher quality sources of pipe (97),
pump (103), and well (102), so a shift in the distribution of
these sources has little effect. The mortality differentials
are somewhat larger for types of sanitation, so the effect of
a shift in distribution is slightly larger than for drinking
water.
The implication of this pattern of mortality
differentials by source of drinking water is that if effort
and expense are expended on providing pipe/tap water to
households which already have wells, then the impact on
mortality may be very small. Much greater impact would be
achieved by providing wells for households who currently rely
on spring, river or rain water.
In conclusion, it has been shown that infant mortality
declined in East Java by some 28 points (99 to 71 per 1,000)
(or 11.2 points, from 91.6 to 80.4 according to the Last Birth
Method) during the period 1980 to 1985. The above estimates
suggest that very little of this substantial decline (0.6-1.3
points) was due to the expansion in high quality drinking
water and sanitation facilities.
What then were the factors which account for the
remainder of these declines? It can only be assumed that
these other factors include improvements in nutrition, in
health services, in safer fertility and delivery practices,
etc., but their individual contributions cannot easily be
quantified.
3.3 Changes in Provincial Distributions of Water Supply and
Sanitation Facilities, Indonesia, 1980 to 1985
This subsection is really incidental to the above
analysis and simply presents the percentage distributions (and
absolute differences) of facilities across the provinces in
both 1980 and 1985.
The data in Table 4 show drinking water facilities which
have seen a moderate increase in proportions of households
with access to ’potable' water. The increase in piped water
supplies to 10.8% of households overall was due mainly to a
large increase in urban areas (1.35 million of the 1.74
million households overall) where now one in three households
have access to piped water. This may not necessarily be piped
into the house itself, but a tap may be available nearby. The
increase of 1.62 million households with pump water supply was
6
equally divided between the urban (8.14 million) and rural
(8.05 million) areas.
The provincial patterns are variable with Lampung having
the lowest proportion of households with piped (1.52%) or pump
water (2.15) in 1985, compared to Jakarta with 34.5% pipe and
45.94% pump water. Of the basically rural provinces, Bali has
19.99% households with piped and 3.48% with pump water, and an
unusually high proportion (29.48%) of spring water - partly
accounted for by the religious beliefs about the spiritual
benefits of spring water, but also because of its plentiful
supply. NTT households (46.93%) and Irian Jaya (28.95%) also
rely heavily on spring water, largely through lack of
alternatives. River water is another major source of drinking
water in Irian Jaya (28.27%), as well as in parts of Sumatera
such as Jambi (29.93%) and South Sumatera (Selatan) (34.55%),
and all of Kalimantan - 50.84% in West (Barat), 72.74% in
Central (Tengah), 41.88% in South (Selatan), and 34.44% in
East (Timur).
The provincial patterns of sanitation facilities show
wide variation. Overall about one in seven (14.90%) of
households have a private toilet with septic tank (38.39% of
urban, and only 6.81% of rural households), and another one in
five have (19.74%) have a private toilet without septic tank,
leaving two-thirds (65.35%) of all households using public,
shared or other sanitation. As might be expected, the more
urban provinces have higher proportions of households with
private toilets (with or without septic tank). Some 61.5% of
households in Jakarta, 50.4% Yogyakarta, 44.9% in North
Sumatera, and surprisingly, 61.4% in Lampung, 58.5% in N.T.
Timur and 48.3% in East Kalimantan.
Probably some caution should be used in interpreting the
category 'private toilet without septic tank' as some
provinces, such as Lampung and N.T. Timur generally have
rather poor facilities. Thus it may not be safe to
extrapolate to other provinces the pattern of lower infant and
child mortality rates seen in East Java for the above
mentioned facility (95/1,000) in comparison to
public/shared/other toilet (111/1,000) (see Table 2).
4,CONCLUSION
The data on trends in drinking water supply and
sanitation facility shows that major efforts are underway to
expand the supply of protected water and more hygienic
toilets. This expansion has managed to keep ahead of the
increase in numbers of households due to population growth.
The pattern of mortality rate according to type of
facility suggests that the trend of shifting households away
from well drinking water to pump and pipe drinking water may
have very little effect on mortality.
If in 1980, for
example, East Java had piped water to 100% rather than just 8%
of households, the overall infant mortality rate would have
been 97 per 1,000 compared to the actual 105 per 1,000 - a
7
slight difference. By far the most widely used source of
water is well, and it may be advantageous to direct further
efforts into upgrading existing unprotected wells by adding
casings or rims, and covers, though some societies prefer
drinking water to be open to sunlight in order that it be
’live’.
Education about the importance of boiling drinking
water must also be a component of the drive to provide safe
supplies to the entire population.
The provision of higher quality sanitation has proceeded
more slowly than that of drinking water. Still two thirds of
the Indonesian population do not have a private toilet but use
’public, shared or some other’ form of sanitation, often a
nearby area of bushland or river. The question is again what
is the best way to proceed? If in 1980, East Java had 100% of
households with private toilet without septic tank, the infant
mortality rate would have been 95 per 1,000 rather than 105.
If all households had private toilet with septic tank - a very
expensive undertaking - the infant mortality rate would have
been 80 per 1,000, a substantial decline on the actual level.
Finally, a word of caution on interpretation.
The
presence of higher quality sources of drinking water and types
of sanitation often reflect higher economic status and
associated literacy levels, hygiene knowledge, etc. Thus the
installation of clean drinking water without hygiene education
on hand washing, etc., may have less impact on mortality than
that suggested by the above analysis.
Indeed this is the
pattern emerging from a recent study in Teknaf, Bangladesh,
(N. Islam, personal communication).
REFERENCES
BAPPENAS 1984 (National Development Planning Agency, Republic
of Indonesia) REPELITA IV The Fourth Five Year
Envelopment Plan of Indonesia, 1984/85-1988/89, May 1984.
Streatfield, K. and A. Larson (1987) ’The Intercensal Survey
of Indonesia: 4. Infant and Child Mortality Levels’,
Child Survival Research Note No. 14CS, 23 July, pp.8.
Acknowledgement
The assistance of Jean Hughes and Pat Mooney in the
preparation of these tables is much appreciated.
8
Table 4:
Distributions of Households by Source of Drinking Water, 1980 and
1985, Indonesia by Province
Province
Pipe
Punp
Well
Spring
River
Rain
Other
N.S.
Total
N
D.I. Aceh X 1980
X 1985
diff
2.77
5.14
16359
1.45
2.00
4416
75.18
75.46
57130
6.75
5.67
-1549
11.03
8.33
-8190
1.26
0.63
-2829
1.51
2.77
8755
0.06
0.00
-324
100
100
73768
530673
604441
Sunatera % 1980
Utara
X 1985
diff
11.26
17.00
131704
2.37
2.77
13159
51.28
50.97
123896
17.02
15.74
19890
14.52
10.34
-38519
1.95
1.10
-10392
1.48
2.08
14574
0.11
0.00
-1653
100 1548323
100 1800982
252659
Suaatera X 1980
X 1985
Barat
diff
4.22
9.10
36222
2.40
5.04
19667
43.36
46.02
28328
31.52
27.86
-20018
13.17
8.64
-30085
3.05
2.85
-802
2.15
0.49
-11598
0.13
0.00
-938
100
100
20776
704010
724786
Riau
X 1980
X 1985
diff
5.25
4.51
1410
0.57
0.96
2574
44.86
50.14
71152
1.12
2.22
6704
21.80
17.06
-2833
25.57
24.60
20224
0.70
0.47
-494
0.12
0.04
-329
100
100
98408
413384
511792
Jaibi
X 1980
X 1985
diff
1.91
4.88
16147
2.94
2.97
4483
40.57
40.55
60031
3.13
3.83
7751
36.52
29.93
24592
14.45
17.24
33934
0.41
0.60
1458
0.06
0.00
-181
100
100
148215
300076
448291
Suaatera X 1980
Selatan X 1985
diff
11.06
14.92
70445
1.64
0.64
-6999
43.80
42.91
99759
3.58
4.19
15699
36.67
34.55
68204
2.28
1.93
1827
0.81
0.79
1793
0.17
0.08
-564
100 857338
100 1107502
250164
Bengkulu X 1980
X 1985
diff
2.87
6.11
7343
1.18
1.10
322
52.41
63.23
41809
13.14
10.27
-151
29.11
18.88
-7729
0.10
0.07
-10
1.11
0.33
-1040
0.09
0.00
-134
100
100
40410
Laipung
X 1980
X 1985
diff
1.44
1.52
6929
1.06
2.15
18328
74.45
81.86
400049
12.11
6.40
-23629
8.38
5.51
-2374
0.22
0.50
4433
2.27
2.03
6217
0.07
0.04
-84
100 871666
100 1281535
409869
Jakarta
X 1980
X 1985
diff
28.19
34.50
287003
31.14
45.94
456825
28.90
14.47
-78383
0.05
0.09
1159
0.30
0
-3453
0.16
0.17
1229
11.22
4.82
-44620
0.06
0.00
-648
100 1164082
100 1783194
619112
Java
Barat
X 1980
X 1985
diff
3.71
7.24
321072
5.51
12.30
593848
58.95
53.99
487410
23.64
6.17
21.41
3.46
176857 -114614
0.09
0.20
9170
1.82
1.35
-9084
0.09
100 6100713
100 7564157
0.06
-1215 1463444
Java
Tengah
X 1980
X 1985
diff
5.47
6.57
65091
2.17
4.68
137548
63.30
63.42
72700
20.37
19.04
-50236
6.13
4.57
-77402
0.26
0.22
-1458
2.16
1.47
-35107
0.14
0.03
-6073
100 5286220
100 5391283
105063
X 1980
D.I.
YogyakartaX 1985
diff
3.36
4.68
11758
1.08
2.64
11405
77.50
77.48
64262
9.35
7.99
-1418
1.34
1.24
401
1.07
3.60
17998
6.19
2.35
-20854
0.12
0.04
-434
100
100
83118
Java
Tiaur
X 1980
X 1985
diff
8.08
12.23
374598
2.26
4.53
185681
67.08
63.08
286655
15.74
14.95
78385
4.53
3.17
-60522
0.37
0.05
-20472
1.83
1.99
27298
0.10
0.01
-5577
100 6478680
100 7344726
866046
Bali
X 1980
X 1985
diff
12.57
19.99
49470
1.91
3.48
9963
35.09
32.95
11821
33.54
29.48
145
13.56
9.72
-12119
0.97
1.93
5945
2.27
2.42
2331
0.09
0.04
-189
100
100
67367
485201
552568
N T Barat % 1980
X 1985
diff
0.99
4.53
24064
3.07
8.29
36517
67.30
59.36
-7939
19.15
19.01
11737
8.72
7.66
-1227
0.00
0
-21
0.72
1.06
2750
0.06
0.08
168
100
100
66049
594192
660241
9
150218
190628
592563
675681
Table 4: (contd.)
Province
Pipe
Punp
Well
Spring
River
Rain
Other
N.S.
Total
N
N T Tiiur X 1980
X 1985
diff
12.95
16.03
27474
0.68
0.67
448
24.16
23.02
11840
47.00
46.93
35303
12.25
10.70
459
1.18
1.06
238
1.26
1.57
2733
0.52
0.02
-2496
100
100
75999
495942
571941
KalinantanX 1980
Barat
X 1985
diff
3.13
4.82
13152
0.08
0.47
2287
19.05
11.13
-23797
2.83
1.61
-3821
49.24
50.84
64484
24.94
30.65
60570
0.63
0.44
-398
0.09
0.06
-90
100
100
112387
458218
570605
KalinantanZ 1980
X 1985
Tengah
diff
0.37
3.67
10023
3.23
7.49
15909
15.92
13.25
9159
2.67
0.38
-3845
76.49
72.74
70621
1.07
2.06
4030
0.13
0.33
741
0.12
0.08
4
100
100
106642
185528
292170
KalinantanX 1980
Selatan X 1985
diff
11.79
17.52
37251
7.04
12.25
31435
29.73
25.85
168
1.23
1.11
254
49.17
41.88
-4185
0.67
1.25
3430
0.23
0.04
-791
0.14
0.08
-204
100
100
67358
444435
511793
KalinantanX 1980
Tinur
X 1985
diff
9.37
19.78
38311
6.11
8.67
12102
30.33
22.60
-2270
3.15
4.56
6518
42.07
34.44
6278
8.59
8.14
4666
0.30
1.76
4665
0.08
0.04
-51
100
100
70219
234557
304776
Sulawesi X 1980
Utara
X 1985
diff
8.92
14.49
40624
1J6
1.87
4442
60.42
58.21
.65183
19.62
20.55
29829
6.22
2.75
-10307
1.16
1.57
3663
1.84
0.51
-4683
0.47
0.04
-1675
100
100
127076
398993
526069
Sulawesi X 1980
X 1985
Tengah
diff
3.86
7.74
14974
6.34
9.37
14244
41.45
44.84
42374
17.80
14.02
1980
28.82
21.35
-1012
0
0.19
-603
1.61
2.49
3947
0.11
0
-263
100
100
76847
233121
309968
Sulawesi X 1980
Selatan X 1985
diff
5.97
11.40
78079
2.27
5.28
41760
68.63
60.61
3185
13.00
13.15
21767
8.80
8.04
3830
0.30
0.03
-3009
0.95
1.44
7673
0.08
0.05
-281
100 117330
100 1270334
153004
Sulawesi X 1980
Tenggara X 1985
diff
2.82
11.97
21019
0.92
2.05
2844
60.59
56.84
17838
19.02
16.09
1798
14.38
11.90
792
2.16
0.43
-2819
0.00
0.72
1550
0.11
0.00
-198
100
100
42824
Maluku
X 1980
X 1985
diff
10.40
11.87
10618
0.73
0.94
1059
55.62
64.33
59221
19.90
12.97
-7909
11.19
9.04
618
1.73
0.66
-2035
0.30
0.18
-177
0.13
0.00
-298
100 228689
100 289786
61097'
Irian
Jaya
X 1980
X 1985
diff
10.19
13.62
13991
1.21
1.65
1746
18.58
19.11
10394
30.90
28.95
9827
32.18
28.27
5252
4.25
6.03
6769
2.65
2.37
557
0.05
0.00
-99
100
100
48437
10.53
17.30
8.54
15.69
395415 -120832
1.50
1.64
134882
2.13
1.68
-40094
100 30263273
0.12
100 35889411
0.03
-23712 5626138
2.11
1.30
-10088
1.41
1.46
46648
3.15
1.42
-63299
0.28
100 6167198
0.02
100 9190151
-14709 3022953
12.68
21.04
11.03
20.27
343066 -110744
1.53
1.71
88234
1.87
1.77
23205
0.07
100 24096075
0.03
100 26699260
-9003 2603185
X 1980
Total
Indonesia X 1985
diff
57.43
3.96
7.03
53.78
7.85
10.77
1619091
1922919
1738469
Urban
X 1980
Indonesia X 1985
diff
26.44
32.45
1351008
11.17
16.35
813817
Rural
X 1980
Indonesia X 1985
diff
2.06
3.31
387461
58.63
2.12
56.94
4.93
805274 1075692
52.73
44.60
847227
2.72
2.40
52349
10
173598
216422
215523
263960
Table 5:
Distributions of Households by Sanitation Facility,
1980 and 1985, Indonesia by Province
Toilet Facility
Private Private Shared/
With S.T No S.T. Public/
Other
N.S.
Total
100
1-00
73768
Province
D.I. Aceh % 1980
X 1985
diff
7.53
12.10
33217
23.02
22.34
12874
69.12
65.56
29466
0.34
0.00
-1789
Suaatera
Utara
X 1980
X 1985
diff
11.41
20.67
195670
33.20
34.44
106298
55.06
44.87
-44330
0.34
0.01
-4979
Suaatera
Barat
X 1980
X 1985
diff
3.99
9.70
42215
7.04
7.46
4496
88.55
82.81
-23190
0.42
0.03
-2745
100
100
20776
Riau
X 1980
X 1985
diff
12.02
19.08
47956
25.67
32.72
61327
61.88
48.17
-9268
0.43
0.04
-1607
100
100
98408
Janbi
X 1980
X 1985
diff
6.31
9.29
22744
16.68
24.38
59236
76.52
66.33
67738
0.50
0.00
-1503
100
100
148215
Suaatera
Selatan
X 1980
X 1985
diff
12.16
14.55
56828
17.80
19.54
63768
69.55
65.92
133793
0.49
0.00
-4225
100
100
250164
Bengkulu
X 1980
X 1985
diff
5.37
10.12
11224
12.68
13.80
7250
81.44
76.09
22699
0.51
0.00
-763
100
100
40410
Laspung
X 1980
X 1985
diff
6.90
15.45
137873
42.28
45.91
219803
50.45
38.64
55398
0.37
0.00
-3205
100
100
409869
Jakarta
X 1980
X 1985
diff
40.37
51.38
446216
11.34
9.11
30557
48.13
39.49
143795
0.16
0.02
-1456
100
100
619112
Java
Barat
X 1980
X 1985
diff
6.74
11.80
481701
7.26
10.34
338953
85.49
77.84
672174
100
0.50
100
0.02
-29384 1463444
Java
Tengah
X 1980
X 1985
diff
6.47
12.24
317764
74.25
18.82
66.57
21.17
146130 -336084
0.45
0.02
-22747
100
100
105063
X 1980
D.I.
Yogyakarta X 1985
diff
11.59
16.46
42589
31.60
33.89
41792
56.45
49.62
748
0.37
0.02
-2011
100
100
83118
Java
Tiaur
X 1980
X 1985
diff
7.89
12.99
442998
21.68
22.72
264057
70.09
64.29
180901
0.34
0.00
-21910
100
100
866046
Bali
X 1980
X 1985
diff
11.81
19.64
51233
6.11
10.03
25812
81.71
70.28
-8088
0.38
0.04
-1590
100
100
67367
‘
11
100
100
252659
Table 5: (contd.)
Toilet Facility
Private Private Shared/
With S.T No S.T. Public/
Other
N.S.
Total
Province
N T Barat X 1980
X 1985
diff
3.20
5.23
15548
3.28
4.39
9497
93.10
90.30
43037
0.43
0.08
-2033
100
100
66049
N T Tiaur Z 1980
% 1985
diff
4.85
6.96
15730
34.77
51.56
122430
59.41
41.48
-57360
0.97
0.00
-4801
100
100
75999
Kalinantan Z 1980
Barat
Z 1985
diff
5.77
10.66
34398
17.26
18.47
26313
76.54
70.84
53484
0.44
0.04
-1808
100
100
112387
Kalinantan Z 1980
Tengah
Z 1985
diff
3.46
6.22
11767
6.67
7.39
9216
89.42
86.39
86498
0.45
0.00
-839
100
100
106642
Kalinantan Z 1980
Selatan
Z 1985
diff
6.39
9.31
19238
12.94
12.75
7751
80.16
77.90
42404
0.50
0.04
-2035
100
100
67358
Kalinantan Z 1980
Tiaur
Z 1985
diff
17.45
27.68
43428
16.45
20.66
24378
65.61
51.60
3360
0.49
0.06
-947
100
100
70219
Sulawesi
Utara
Z 1980
Z 1985
diff
14.08
19.94
48720
21.24
24.95
46479
63.85
55.11
35191
0.83
0.00
-3314
100
100
127076
Sulawesi
Tengah
Z 1980
Z 1985
diff
7.55
10.04
13507
13.53
16.13
18474
78.44
73.83
45984
0.48
0.00
-1118
100
100
76847
Sulawesi
Selatan
Z 1980
Z 1985
diff
8.58
12.29
60205
16.83
17.96
40064
74.04
69.76
58902
0.55
0.00
-6167
100
100
153004
Sulawesi
Tenggara
Z 1980
Z 1985
diff
4.49
8.20
9957
34.62
28.80
2217
60.52
63.00
31295
0.37
0.00
-645
100
100
42824
Maluku
Z 1980
Z 1985
diff
9.06
10.37
9329
5.63
4.85
1164
84.85
84.72
51481
0.45
0.06
-877
100
100
61097
Irian
Jaya
Z 1980
Z 1985
diff
7.76
15.26
23544
17.36
10.01
-11005
74.40
74.74
36918
0.47
0.00
-1020
100
100
48437
Z 1980
Total
Indonesia Z 1985
diff
17.74
8.94
72.89
0.43
100
14.90
19.74
0.01
100
65.35
2640387 1716609 1394660 -125518 5626138
Z 1980
Urban
Indonesia Z 1985
diff
28.88
38.39
1746560
17.38
17.04
493848
53.30
44.57
809014
0.44
100
100
0.01
-26469 3022953
Rural
% 1980
Indonesia % 1985
diff
3.84
17.83
20.67
6.81
893827 1222761
77.90
72.50
585646
100
0.43
100
0.02
-99049 2603185
,
12
A-. <
Research Note on
CHILD
SURVIVAL
Number
18CS
Date
23 June 1988
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
CHILD MORBIDITY PATTERNS IN ETHIOPIA, 1983
Aynalem G. Yohannes
Central Statistics Office,
Addis Ababa, Ethiopia
and
Kim Streatfield
Child Survival Project,
International Population Dynamics Program,
Department of Demography,
The Australian National University
Note:
Child
Survival Research Notes are brief discussions of
issues of current relevance to researchers and policy
makers concerned with Problems of high infant and child
mortality in the world. The International Population
Dynamics Program, Department of Demography, The Australian
National University, distributes these notes with their
regular Bibliographic Circular. Production of the
Child Survival Research Notes is made possible through
a grant from the Ford Foundation (840-0893)Responsibility for the content of Child Survival
Research Notes rests with the author(s) alone, and not •
J;he above-listed organisations.
There is a widespread impression that Ethiopian children
are in a chronic state of marginal health interspersed with
periodic disasters, such as famines. Until recently, reliable
data on the levels and patterns of child health (or ill
health) has been scarce, but the Central Statistical Office
conducted a brief very useful health survey of the rural areas
of Ethiopia in 1983. The resulting data is examined here for
patterns of child illness. A subsequent Child Survival
Research Note will examine the utilization of the traditional,
and the relatively limited modern, health services.
BACKGROUND:
A study carried out by the Ministry of Health (1980: 20)
indicated that about 70 percent of the all deaths in Ethiopia
were among children under the age of 5. UNICEF estimated in
1986 that one quarter of all children died before reaching
their fifth birthday (UNICEF, 1988:64).
The infant mortality
rate was high at 151 per 1,000 live births, and some 60% of
under fives were suffering from mild or moderate malnutrition
(10% from severe malnutrition) (ibid).
Based on the first
Morbidity Survey in 1982, the Central Statistical Office (CSO)
(1985: 40) found that abdominal diseases, which include
diarrhoea, gastroenteritis, parasites and pain in the stomach,
have been the most serious health problems.
Health interventions have played an important role in
child survival, especially in preventable diseases through
immunization, provision of clean drinking water and improved
sanitation, and access to health services (United Nations,
1982). However, in Ethiopia, immunization services are not
widespread and ante- and post-natal services are not adequate.
(1988:68) that in 1985-86 the percentages of
UNICEF report
immunized 1-2 year old children were still very low for TB
(12%), DPT (6%), Polio (6%), and Measles (9%). Compared with
most other developing countries, these are the lowest levels
of immunization coverage.
It appears that in Ethiopia breastfeeding is continued
for long durations. High percentages, 97 and 95 percent, of
mothers were breastfeeding their children for 6 and 12 months
respectively (UNICEF, 1988).
However, according to the World
Health Organization (1981: 131) report, about 14 percent of
these children aged 12 months were still being breastfed with
no other supplementary food given.
From other studies, such
as in India, it has been observed that late introduction of
supplementary food may lead to a lower rate of growth and
other health problems (ibid: 141).
SOURCE.QF _ DATA:
The study uses data from the Rural Health Survey of
Ethiopia that was undertaken during late January 1983.
This
survey was one of two rounds of health surveys, which were
conducted during 1982-83 by the Central Statistical Office as
part of a Rural Integrated Household Survey Programme (RIHSP).
1
In this survey, all members of the selected households
who were present during the interview (except children) were
asked by enumerators (non-medical persons) to give information
on some socio-demographic characteristics, perceived morbidity
or injury during the 14 days prior to the survey date, and
source of treatment .
The morbidity information was obtained
by asking whether or not the respondent was ill during the
reference period.
If the respondents reported they were ill,
then a further question was asked about the types of disease
they suffered from.
To make sure that the reported disease
classification was correct, enumerators were told to ask the
symptoms and compare them with their supplied list of symptoms
and their diseases category.
In this study 11,962 children of the heads of the
household under age six have been selected.
Along with each
child’s data, some information on the child’s mother and the
household sanitary conditions was included for the analysis,
to examine the relationship between maternal characteristics
and the child’s morbidity.
The selected maternal variables
include information on socio-demographic characteristics (e.g.
age, literacy, religion, ethnic group) and morbidity of the
mother; while the household sanitary variables include
information availability of pit-latrine, refuse disposal
(garbage bin) and drinking water supply.
The 1983 Rural Health Survey was based on a two-stage
stratified sample design.
This survey covered the rural
sedentary population of twelve of the fourteen administrative
regions of Ethiopia, excluding Eritrea and Tigrai regions.
Stratification was carried out for each of the twelve regions
at sub-regional level.
The sub-regions (awrajas) were
adopted as the first level of stratification, and thus 77
units were formed.
Within each unit the Farmers’
Associations (FAs), used as the primary sampling unit (PSU),
were selected with probability proportional to the size of the
administrative regions.
In total, this yielded a sample of
500 FAs to be selected from the twelve regions.
However,
because of inaccessibility and other reasons the survey
actually covered only 476 FAs.
A complete listing of all households in each of the
selected PSUs was prepared.
Using simple random sampling, 25
agricultural and 1 to 5 non-agricultural households were
selected from the list of each PSU.
All members of the
selected households were eligible for interview.
.... _s.u_r y sy
Awareness of illness-. Among different societies, there
are differentials in perception of illness.
It is observed
that some societies do not consider even diseases that are
major causes of death as illness.
Robinson (1971, cited in
Greenley, 1980: 178) stated that chronic conditions are not
reported by some individuals, especially among illiterates
because they are not seen as unusual.
In this survey an
2
attempt was made to reduce enumerator bias by providing a list
of symptoms of different diseases for the question on sickness
in the last 14 days.
Memory recall: Another problem which is likely to affect
the quality of data is the mis-reporting of illness due to
’memory lapse’.
The length of reference period is important
in collecting retrospective data.
Severity of illness: This health survey attempted to
examine the severity of illness by asking 'days of restricted
activity’ for the symptom or condition reported.
FINDINGS:
Age-sex di f f erent lais in chi.ld_. morbid.it.y.
The data in Table 1 show the age-sex specific morbidity
rates of children. The sex differentials fluctuate from age
to age and no clear pattern is observed.
For children aged
1, 4, and 5 years, morbidity rates are found to be slightly
higher among females than males.
But overall, the difference
is seen to be small, about 0.2 percent, and statistically
insignificant.
This conforms to findings of other Ethiopian
studies of mortality, completed in three regions, which show
no significant sex differentials (Genet, 1987; Kebede, 1986).
For both males and females, the morbidity rate reaches its
peak at age 1, with a level of 35 percent for males and 36
percent for females, and then decreases as the children grow
up.
This observed differential could be due to exogenous
factors, such as increased exposure to contaminated weaning
foods in the second year of life, and due to the immune
system, that is, infants and toddlers are more likely to have
weaker immune systems than older children (Johansson and Mosk,
1987: 213).
Table 1:
Age in
Years,..
0
1
2
3
4
Total
Note :
Percentage of Sick Children by Age and Sex
Male
34.1
34.7
31.9
27.5
22.3
Differences in Percentage
...Female....... ..... .. Male-.Female
31.8 (1061)
36.2 (1061)
29.9 (1014)
24.6 (1010)
25.3 ( 891)
_ 9.3 -Dj....23^J 930)
(1088)
( 969)
(1013)
(1060)
( 972)
28.9 (6037)
28.7 (5925)
2.3
-1.5
2.0
2.9
-3.0
-1...:.4____
0.2
Numbers in brackets are the base numbers for the
percentages.
3
The pattern of sickness in relation to age may differ
according to type of disease.
Therefore, the prevalence
rates of the selected 11 major diseases and the residual
group, ’other diseases’, are shown in Tables 2 and 3, for male
and female children respectively.
Among all ill children,
these tables show that diarrhoea and gastro-enteritis are the
most prevalent.
This conforms with observations all over the
world, where diarrhoea is found to be the most prevalent
childhood disease (Black et al., 1983: 141).
Table 2:
Prevalence Rate (per thousand) of Diseases Among
Male Children by Age
,Type„of Illness.
...........
4
5
.-..T.Q.ta.l
(A)
Diarrhoea &
Castro-Enteritis
Parasitic
95.6 132.1
7.4 20.6
90.8
22.7
49.1
28.3
36.0
19.5
29.9
26.7
72.7
20.7
(B)
Whooping Cough
Pneumonia
Other Respiratory
20.2
4.6
68.9
16.5
5.2
48.5
9.9
3.0
66.1
8.5
2.8
58.5
8.2
4.1
37.0
6.4
0.0
50.3
11.8
3.3
55.3
(C)
Eye Infection
Skin Infection
Malnutrition
Malaria
Measles
Tetanus
Q..^'.bG.r D_i_s eases
39.5
28.5
2.8
2.8
5.5
0.0
65^3
19.6
35.1
9.3
3.1
6.2
1.0
49,5
20.7
23.7
11.8
4.9
3.9
0.0
61.2
26.4
9.3
24.5 30.8
6.6
9.3
4.1
2.8
8.2
5.7
0.0
0.0
6Q..?,4.. 56,6
16.0
32.1
1.1
3.2
7.5
0.0
22.4
29.0
6.8
3.5
6.1
0.2
57x0
Total
341.0 346.7 318.9 273.6 223.3 220.3
288.7
Number of Cases
Note :
1088
(A) is the group
(B) is the group
(C) is the group
groups A and
969
1013
1060
972
935
6037
of abdominal diseases,
of respiratory diseases.
of diseases other than in
B.
The data in Table 2 indicate that about one in fourteen
(73 per thousand) male children were ill in the last 14 days
with diarrhoea and gastro-enteritis.
It appears from this
table that male children under age 2 experience this disease
more than those who are older.
This may be explained by
contaminated supplementary foods introduced at weaning age
(Black et al., 1982: 259-264).
After the second birthday, the prevalence of diarrhoeal
disease declines steadily with age.
This finding is
4
consistent with the study in Bangladesh that reveals incidence
of diarrhoeal disease to be high in the first two years of
life (Black et al., 1983: 80-83).
In addition, a significant
shift in the leading cause of illness, from diarrhoea to
’other respiratory’, is observed after the first three years
of life.
For all children under six years, the next most prevalent
disease, with a rate of 55.3 per thousand male children, is
’other respiratory’.
This category includes a very small
percentage of tuberculosis (about 0.3 percent), and a large
percentage of coughs and colds.
The rates for this specific
type of illness fluctuate from age to age and seem to reach
their lowest level at age 4.
Surprisingly, the level of malnutrition was found to be
very low, about 7 per thousand, compared to the findings of
other reports in Ethiopia.
As mentioned above, according to
UNICEF, in 1980-86 about 10 percent of children in Ethiopia
were severely malnourished, and 60% mild to moderately
malnourished.
In the current survey, however, the children
were not weighed, but nutritional status was based only on
visual assessment by the enumerators.
The low level of
malnutrition found by this study could also be because this
condition is often associated with health conditions such as
infections, hence respondents might have reported the
consequential disease as important rather than the underlying
malnutrition.
The prevalence rates of the selected diseases for female
children are given in Table 3.
The distribution pattern
(regardless of age) of these diseases is found to be similar
to that for males except for malaria, measles and
malnutrition.
For instance, as in the case of males,
prevalence of diarrhoea and gastro-enteritis is the highest
overall, at 80 per thousand.
Moreover, for children less
than 2 years of age, the diarrhoeal morbidity pattern of males
also holds for females.
However, in general, the age
specific morbidity rates do not show as consistent a pattern
as that for males.
Even though the morbidity rates for most
diseases are higher for females than males, the differences
observed are minimal except for diarrhoea and gastro
enteritis .
The residual 'other diseases’ category has also shown
high rates for both males and females.
But this category is
classified broadly, including diseases such as fever, typhus,
meningitis, infectious jaundice, goitre, throat and ear
infections, and teeth and gum problems.
5
Table 3:
Prevalence Rate (per thousand) of Diseases Among
Female Children by Age
. 11. ..
..Type.of....Illness...
____ 3_„
(A)
Diarrhoea &
Castro-Enteritis
Parasitic
105.6 135.4
9.4 19.6
80.9
21.7
49.5
27.7
59.5
24.7
44.1
26.9
80.3
21.4
(B)
Whooping Cough
Pneumonia
Other Respiratory
15.1
3.8
53.7
16.7
3.9
60.8
8.9
1.0
67.1
12.9
4.0
54.5
6.7
2.2
53.9
6.5'
1.1
58.1
11.3
2.7
58.1
45.2
31.1
3.8
1.9
8.5
32.4 18.7 12.9 16.8
25.5 24.7 24.8 23.6
7.9
3.0
5.9
6.-7
2.9
4.9
5.9
9.0
8.8
8.9
3.0
5.6
ALi.2_-58^2__44.,__6.._41..;.5_
16.1
28.0
2.2
6.5
4.3
24.1
26,3
4.9
5.1
6.6
44.9
Total
317.6 360.2 297.8 245.5 250.3 233.3
285.7
Number of Cases
1061
5925
(C)
Eye Infection
Skin Infection
Malnutrition
Malaria
Measles
Other Diseases
Note:
..
1019
1014
1010
891
930
(A) is the group of abdominal diseases,
(B) is the group of respiratory diseases,
(C) is the group of diseases other than in
groups A and B.
Regional differentials in. child morbidity.
The rate of sickness may differ according to different
geographical areas; this is attributed to variation in
climates, environmental setting, and socio-economic factors.
The data in Table 4 display the age standardized morbidity
rates by sex of children for each of the twelve regions. The
rates are standardized to remove any differences due to
variations in age distributions in each area. Most of the
rates are above 20 percent.
In four regions, Conder,
Illubabor, Wellega and Kefa, the morbidity rates greatly
exceed that of the country as a whole, which is about 29 per
cent.
The first three regions mentioned are found to have
the highest morbidity rates with 59, 46 and 42 percent
respectively.
Except in Shewa, which has a significantly
lower rate (19%), the rest vary from 21 to 29 percent.
It
appears from this table that there is no significant sex
differential in most of the regions.
The largest gap is
found in Gojam and Hararge with about 5 percent of excess
female morbidity.
6
Table 4:
Age Standardized Morbidity Rates of
Children by Region and Sex
.'1.
*-i-■
.
,Male„..
Arssi
Bale
Gamo-Gofa
Go jam
Gonder
Hararge
Illubabor
Kef a
Shewa
Sidamo
Wellega
Wollo___
24.9
21.3
22.9
20.2
57.7
18.9
42.4
30.9
19.5
22.0
46.3
...„2Q_._2.
23.5
22.9
21.5
24.7
60.7
23.6
41.0
27.3
18.4
21.2
45.6
.. .21
24.1 ( 1086)
21.8 ( 851)
21.9 ( 733)
22.2 ( 949)
59.3 ( 980)
21.5 ( 897)
41.9 ( 711)
29.1 ( 814)
18.9 ( 2181)
21.6 ( 783)
45.7 ( 1303)
..20<.._5._t__(_. 674).
Total
28.9
28.7
28.8 (11962)
Note:
Direct standardization is used; that is
the population of the whole country is
taken as standard.
The prevalence of different types of disease according to
2Table
“ * ",
It can be seen from
these regions is revealed in
5.
this table that the iprevalence of diarrhoea accounts for the
Gonder.
highest level of morbidity found in Gonder.
The Central
Statistical Office (1985) report confirms that in 1983 when
the survey was being carried out, an epidemic of diarrhoeal
disease occurred in this region.
Moreover, it can also be
seen that malnutrition is highest in this region.
As
mentioned above, deficient food intake has an association with
diseases such as infections, thus this could be one of the
explanations for most of the diseases occurring more in this
region than the others. Similarly, in Wellega the level of
malnutrition is found to be high.
Hence, this poor nutrition
may contribute to some extent to the observed high morbidity
in Wellega region.
In
In Illubabor,
Illubabor, ’other respiratory’ and ’parasitic' are two
extremely high prevalence diseases, which account for the
observed high morbidity level.
On the other hand, in Shewa,
the high concentration of health services could be the reason
— ■ morbidity
....
People
in this
for the observed lowest child
rate.
f
region are more likely
1--- - than people in other regions to be
aware and seek modern health services because they are closer
to the capital city, Addis Ababa, which has the highest
concentration of health practitioners_att one doctor per 5,789
•
for the whole country,
population, compared• to
one -per- 57,867
J
and one per 312,000 population in Gamo-Gofa (Yohannes,
1987:12).
7
Table 5:
Prevalence Rate (per thousand) of Specific Diseases by Region.
Types of
Diseases
Arssi
Bale „Gofa___ Go jam _Gonder_ Harage
Diarrhoea &
Gastro-enteritis
47.0
56.4
58.7
54.8
148.0
Parasitic
30.4
1.2
4.1
8.4
Whooping Cough
4.6
18.8
9.5
Pneumonia
5.5
0.0
Other Respiratory 57.1
Eye Infection
Camo-
Illubabor
Kefa_.
Shewa
sidamo.. Wellega
JW0I.I.0
52.4
98.5
78.6
54.6
47.3
149.7
65.3
37.8
30.1
81.6
16.0
7.3
7.7
35.3
5.9
6.3
12.2
11 . 1
7.0
19.7
13.3
17.9
9.2
8.9
0.0
4.2
3.1
0.0
1.4
8.6
2.3
2.6
4.6
3.0
36.4
34.1
51.6
78.6
44.6
105.5
65.1
42.6
60.0
84.4
23.7
19.3
15.3
15.0
23.2
76.5
8.9
18.3
16.0
12.4
3.8
39.9
29.7
Skin Infection
25.8
37.6
15.0
19.0
89.8
12.3
29.5
25.8
11.9
19.2
35.3
20.8
Malnutrition
0.0
2.4
5.5
5.3
15.3
4.5
9.8
7.4
2.3
2.6
13.8
3.0
Malaria
2.8
0.0
8.2
1.1
23.5
0.0
9.8
0.0
1.4
7.7
0.0
3.0
Measles
1.8
7.1
5.5
0.0
5.1
7.8
4.2
13.5
6.0
2.6
15.3
4.5
Qther.Diseases
46.0 ___4X.5
61.A
45.^3
..l.O^.x.l____
—.. 50;6____ 4£.1.8...„„_32^1.
_ 4_3^_4
Number of Cases
1086
851
733
949
980
1303
674
897
711
814
2181
783
Morbidity Patterns According.
Several studies have revealed an association between
environmental sanitation and health status.
In Keneba,
Gambia, Rowland (1983:94-95) observed the association between
poor sanitation and seasonal peaks in diarrhoea caused by high
levels of food contamination.
He found diarrhoeal disease to
be prevalent during rainy seasons because of weaning foods
being prepared from well water.
Hobcraft et_al. (1984:??)
assessed the clear relationship between access to toilet
facilities and child survival in Sri Lanka and Mexico.
On the
other hand, Feachem et al. (1983) found that having good
sanitation in individual homes will not protect members of
families unless the overall community environmental sanitation
and faecal contamination is low.
Okediji (1975:293) also
concluded that unhygienic neighbourhoods or environments can
lead to epidemics of diarrhoeal diseases.
Table 6:
Percentage of Sick Children by Sanitary Facilities
(a)
Sanitary
Facilities
(b)
Sick
Children
Stand
ardized
(d)
Diarrhoea
& GastroEnteritis..
(e)
Other
Illness
(C)
_
Water Supply
Pipe
Lake
Well
Spring
River
Other
Chi-square=53.81,
12.7 (
166)
347)
21.0 (
826)
25.7 (
29.8 ( 6087)
30.0 ( 4152)
20.2
356)
d.f.=5, p=0.00
12.7
21.1
25.7
29.8
30.2
20.0
3.6
4.1
6.4
8.1
8.0
5.1
9.1
16.9
19.3
21.7
22.0
15.1
Latrine
Yes
No
Chi-square=23.42,
22.2 ( 1044)
29.4 (10868)
d.f.=1, p=0.00
22.3
29.4
5.8
7.8
16.4
21.6
Usually Defecate
In the latrine
Far from houses
In the back yard
Other
Chi-square=84.22,
488)
17.0 (
23.6 ( 2225)
30.6 ( 8274)
854)
33.4 (
d.f.=3, p=0.00
17.3
23.4
30.7
33.5
5.1
5.3
8.1
11.0
11.9
18.3
22.5
22.4
Garbage Bin
Yes
No
c.h.i~s_quare_—_2_5.58j
17.1 (
380)
29.2 (11531)
d...r_f....=.l_x. P.=.0.....00_.
17.2
29.2
5.3
7.7
11.8
21.5
Notes:
The number of cases did not add up to 11962 due
to some non-response on sanitary facility questions.
2. Numbers in the brackets are base numbers for the
percentage.
3. (b)=(d)+(e).
1 .
8
The data in Table 6 show the relationships between child
morbidity rates and the four variables that are taken as
indicators of sanitary facilities. The data were standardized
for literacy of mother as an attempt to control for economic
status, but there was only a weak relationship between
literacy and the presence of these facilities.
Children living in households which obtain their water
from rivers and springs are more likely to be sick than
children in households having piped water, This is the
expected pattern since piped water is normally assumed to be
less contaminated than other sources.
The presence in a household of a latrine does not
necessarily mean that a child uses it, thus there is a second
indicator used here: 'place where child usually defecates’.
The morbidity rate is lower among children from houses where
there is a latrine, and, as expected, morbidity rates
according to place of defecation are lowest where children use
a latrine in the household, rather than away from the house or
in the backyard.
The final indicator of environmental contamination used
here is the presence or absence of a garbage bin in the house.
Where a bin is present, the morbidity rates are substantially
lower than where there is no bin.
This presumably reflects a
difference in parental attitudes towards cleanliness, rather
than simply the function of the bin itself. A statistically
significant association (based on chi-square test) existed
between all four indicators and the rates of child morbidity.
It is interesting to note that poor sanitary facilities
measured by these four variables attributed more to 'other
diseases' than 'diarrhoeal and gastro-enteric diseases'. This
may be because diseases such as skin infections, eye
infections, parasitic infestations, are also more common under
conditions of poor sanitation, and may be spread by insect
vectors such as flies.
CONCLUSION
In general, the level of child morbidity observed is
quite high, with the morbidity rate at its highest level for
children aged one year.
There do not appear to be any
substantial morbidity differentials between male and female
children.
Sanitary facilities show a strong association with child
morbidity, suggesting that the Ethiopian Government’s Primary
Health Care program, carried out by the Community Health
Services Department, should emphasize improvements in
sanitation.
In addition, since diarrhoea is the most
prevalent disease, mothers should be given health education so
they are familiar with the simple prepared treatment, Oral
Rehydration Salts, as well as having some knowledge about
‘personal hygiene’ especially in preparing supplementary foods
for children.
9
REFERENCES
Black, R.E., K.H. Brown, S. Becker, A.R.M.A. Alim and
M.H. Merson, 1982, "Contamination of Weaning Foods and
Transmission of Entrotoxigenic Escherichia Coli Diarrhoea
in Children in Rural Bangladesh", .Txansactipns. pf_„.Rpy.al.
Society.p f... Txopi c a 1_.Medicine.and. Hygiene , 76: 259-264.
, M.H. Merson and K.H. Brown, 1983,
"Epidemiological Aspects of Diarrhea Associated With
Known Enteropathogens in Rural Bangladesh”, in Diarxhea
and..Malnutrition, eds., L.C. Chen and M.S. Scrimshaw,
pp. 73-86. New York: Plenum Press.
Central Statistical Office, 1985, Report.on.theRural„,Hea2Xh
Survey.. 1982/83 , V o 1.1 , Statistical Bulletin 47.
Addis Ababa: ENI.
Feachem, R.G., M.W.
M. W. Guy, S. Harrison, K.O. Iwugo, T. Marshall,
N. Mbere, R. Muller and A.M. Wright, 1983, "Excreta
Disposal Facilities and Intestinal Parasitism in; Urban
Africa: Preliminary Studies in Botswana, Ghana and
Zambia” , TransacLion^.p„f the Royal. Society of .Tropical
Medicine a nd... Hygiene, 77(4): 515-521.
Genet, M. , 1987, F.e.r..Uli.ty.„.a.n.d Chi.ldhppd.....^
inJural.
Ethippiaj. A Comparative Study. of Selected ...Regions Gpnder
and Hararge, M.A. thesis, Canberra: Australian National
University.
Greenley, J.R., 1980, "Sociocultural and Psychological Aspects
of the Utilization of Health Services”, in Assessing^JtJhe
Cpntributipns. of. the.Social. Sciences. to. Health, eds. M.H.
Brenner, A. Mooney and T.J. Nagy, pp.169-207. USA:
Westview Press.
Hobcraft, J.N. J.W. McDonald and S.O. Rustein, 1984, ’’Socioeconomic Factors in Infant and Child Mortality: A CrossNational Comparison”, Population.Studies, 38(2): 193-223.
Johansson, S.R. and C. Mosk, 1987, "Exposure, Resistance and
Life Expectancy: Disease and Death during the Economic
Development of Japan, 1900-1960” , Population ,.Studies.,
41: 207-235.
Kebede , S. , 1986 , Fer ti.11 ty and.... Chi Id Mortality. in
A.gr.ic.ul...tural Househplds. of Ru.r.a.l....„.Et..h.i.p.p..i..a..; The Case of
Arssi....Adm.i.ni.s.tra.t..i
Region, M . A . thesis , Canberra .Australian National University.
Ministry of Health, 1980, Heal thJ
Addis Ababa: MOH.
10
Study.: „.....Et hippi a ,
Okediji, F.O., 1975, ’’Socioeconomic Status and Attitudes to
Public Health Problems in the Western State: A Case Study
of Ibadan", in PppMlation.„.GroKth„_a
Changes in West Africa, ed. J.C. Caldwell, pp.275-297.
New York: Population Council.
Rowland, M.G.M., 1983, ’’Epidemiology of Childhood Diarrhea in
the Gambia”, in Piarrhea....and.„KaJnutxlti.on eds. L.C. Chen
and N.S. Scrimshaw, pp.87-98. New York: Plenum Press.
UNICEF, 1988 The
UNICEF.
t at e_....of t he WoX1 d s C h i.l d r e n. New York:
Yohannes, A.G., 1988, ’’Differentials in Child Morbidity and
Utilization of Health Facilities in Rural Ethiopia”,
unpublished M.A.(D) thesis. Canberra: National Centre for
Development Studies, Australian National University.
- r-
; •?
*i :
11
Research Note on
CHILD
SURVIVAL
Number
19CS
Date
21 July 1988
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
UTILIZATION OF HEALTH FACILITIES FOR CHILD ILLNESS
IN ETHIOPIA, 1983
Aynalem G. Yohannes
Central Statistics Office,
Addis Ababa, Ethiopia
and
Kim Streatfield
Child Survival Project,
International Population Dynamics Program,
Department of Demography,
The Australian National University
Note:
u (
Child Survival Research Notes are brief discussions of
issues of current relevance to researchers and policy
makers concerned with problems of high infant and child
mortality in the world. The International Population
J&Arnamics Program, Department of Demography, The Australian
$wtional University, distributes these notes with their
Tejrular Bibliographic Circular. Production of the
Cffl-ld Survival Research Notes is made possible through
dFgrant from the Ford Foundation (840-0893).
responsibility for the content of Child Survival
Research Notes rests with the author(s) alone, and not
the above-listed organisations.
INTBQMCTION:
In the previous Child Survival Research Note (No.18) (Yohannes and
Streatfield, 1988) patterns of child morbidity in Ethiopia were examined.
It was seen that reported morbidity levels were highest lor one year old
children, with little difference between boys and girls. Among the younger
children, diarrhoeal disease was the most widespread cause of morbidity,
with respiratory diseases becoming more important among older children (up
to age six). There was considerable regional variation in morbidity
levels.
In this Research Note the focus is on the patterns of use of both
modern and traditional health services when the children are sick. Many of
the patterns of use will be analysed separately for boys and girls.
SQURCE._QF.DATA:
As described in detail in the previous Research Note, these data were
obtained from the Rural Health Survey of Ethiopia carried out by the
Central Statistical Office in January 1983. The survey involved a twostage stratified sample design of twelve of the fourteen administrative
regions of Ethiopia. While the survey inquired about the health of persons
of all ages in the sample households, it included 11,962 children (under
age six years) of the heads of households. Morbidity information was
obtained by asking whether or not the respondent or the child of the
respondent was ill during the past two weeks. If the respondents reported
an illness during the reference period, they were then asked about the type
of illness, including the symptoms experienced. The enumerators were to
compare the symptoms with lists of symptoms and their disease categories.
There is a discussion of some of the pitfalls of such an approach in the
previous Research Note (p.2).
FINDINGS:
Age-sex,.jdifferejnt^...in.jitUization_of.. health.. services
Utilizing health facilities is considered as 'personal illness
control’ behaviour, one of the proximate determinants of child survival
(Mosley and Chen, 1984).
On one hand this variable has an impact on
health status, that is, if there are variations in utilizing health
services, particularly in use of preventive care, there will be health
heterogenity (Rosenzweig and Schultz, 1983:743).
On the other hand,
studies show that using different health facilities can be affected by
access and the socio-economic characteristics of the individual or the
household.
In Matlab, Bangladesh, utilization levels have been analysed among
males and females for all age groups. There was found to be a marked
difference in utilization rates between males and females, with average
treatment rates being 135.6 and 81.9 per thousand respectively among male
and female children under age five suffering from diarrhoea (Chen et al.,
1981:64-65).
In this study a different pattern of utilizing facilities was
observed among male and female children.
Among all sick children under
age 6 (all illnesses) about 33 and 34 per cent of males and females,
respectively, received treatment from at least one of the facilities (Table
1
Of these children who received treatment, most were taken to
not shown).
medical facilities, which include hospitals, health stations and health
there is little difference according to the sex of
centers (see Table 1).
the child in usage of medical facilities.
Table 1:
Percentage of Treated Sick Children
by Type of Facility and Sex
Type of
Facility
Female
40.2
15.1
20.7
17.9
4.9
3.7
(575)
Medical Facilities
Pharmacy
Traditional Healer
Lay Treatment
Self Treatment
Unqualified Person
Number of Cases
Note:
41.4
15.5
18.6
15.5
6.1
5.2
(575)
1. The number of cases is the total number
of children who received treatment from
one of the facilities.
2. Respondents could mention more than one
facility.
v
3. Medical facilities include hospitals,
health centers and health institutions.
‘Traditional healer' is the next most frequent facility among both male and
‘Self treatment’ and ‘Unqualified person' are found to
female children,
It also appears that female children are slightly more
be the least used.
frequent users of modern services (in this study these include medical
facilities and pharmacies) than males.
Table 2:
Percentage Distribution of Sick Children
Who Received Treatment from Medical
Facilities by Age and Sex
Type of Disease
&.Age.in.Years_
Male
.Female
Abdominal Diseases
0
1-2
3-5
22.3 (112)
16.3 (263)
7.9 (189)
18.0 (122)
21.0 (262)
8.7 (219)
Respiratory Diseases
0
1-2
3-5
15.7 (102)
10.8 (148)
8.6 (175)
16.9 ( 77)
15.5 (161)
3.2 (189)
Note:
Abdominal Diseases include diarrhoea, gastro
enteritis and parasitic disease.
2
Since utilization of different treatments depends on the type of
disease, it is plausible to look at the utilization rate of those who were
sick with common diseases.
Therefore, the utilization rates are shown for
the two groups of disease that are most prevalent.
Table 2 presents the
percentage of children who were sick with abdominal diseases and
respiratory diseases and who received treatment from medical facilities.
For children who were exposed to abdominal diseases and respiratory
diseases the medical facilities utilization rate decreases as age
increases.
Male children who suffer from abdominal diseases at infancy
are more likely to be users of medical facilities than females.
This can
be due to severity of illness rather than sex discrimination in using a
medical facility.
Table 3 shows that more male infants (aged under 1
year) who suffered from these diseases were unable to carry out their usual
activities for longer than 8 days compared to females in the same category.
Table 3:
Percentage of Sick Children with Abdominal Disease by
Number of Days Restricted From Activity and by Age
and Sex
.i_n_y.€ar s........
Number of
. D.a.y.s. ..
.Male..
..Q_____
Female.
0
1-7
8+
34.8
40.1
25.0
38.5
45.9
15.6
___ 1-2____________ 3-5
Female
...^1.1.^^...... . . . . . .
■■.
a^aa^^a-
35.7
48.3
16.0
31.3
52.3
16.4
55.0
35.5
9.5
50.2
39.3
10.5
For the above finding, a plausible explanation could be that because male
infants were more severely sick than females then the degree of seeking
treatment was also found to be higher.
This severity of illness could be
due to biological factors.
In infancy boys may be more easily affected by
infectious diseases, or variation in the immune systems of individuals
(Waldron, 1983; Johansson and Mosk, 1987).
The inverse situation occurs
for children aged 1-5.
In Ethiopia, it is believed that the impact of physicians is very
little (only one doctor per 58,000 population), since people usually use
traditional medicine, and even people who are said to have a more positive
attitude towards modern treatment usually buy tablets from pharmacies
without a physician's prescription.
Thus, in the 1983 Rural Health
Survey, reasons were asked for those who did not seek medical facilities
but received only treatment from other facilities, including those who
received treatment from a pharmacy.
Table 4 displays the percentage
distribution of male and female children by the reason they received
treatment only from other than medical facilities.
The most frequent
reason found among both males and females is economic.
However, this
reason is reported slightly more for females than for males, but cultural
reasons, such as using traditional medicine, or not believing that medical
treatment can help, are reported more frequently for male than female
children.
3
Table 4:
Percentage Distribution of Sick Children
Who Received Treatment From Other Than
Medical Facilities by Reason and Sex
Too Expensive
Used Traditional
Medicine
Not Believe it Helps
Disease Will Stop
by Itself
Too Far
Long Waiting Time
Couldn’t Get Required
Medicine
Other
Number of Cases
Note:
25.0
27.6
15.1
9.0
10.6
7.1
21.4
12.3
3.9
27.0
11.5
3.7
3.3
9.9
(332)
4.3
8.1
(322)
The number of cases is the number of sick
children who received treatment from other
than medical facilities.
In summary, differentials.in overall child morbidity according to age
have been observed, levels being higher for infants and toddlers than
children of older ages.
However, no sex differences were found either in
morbidity or in health care practices.
Maternal .„Characteri5tics...and. Child_Morbidity
Literacy Status of Mother
Education is one of the socio-economic determinants that has a strong
effect on child survival.
A number of studies in developing countries
indicate a relationship between parental education, measured either by
literacy status or level of education, and infant and child mortality.
Data from Africa (Caldwell, 1979; Farah and Preston, 1982; Kune, 1980),
Asia (Martin et al., 1983; Cochrane et al., 1980) and Latin America
(Cochrane et al., 1980) show the negative effect of maternal education on
the level of child mortality, although the level of education required to
produce a significant reduction in child mortality varies between
countries.
In the Philippines and Pakistan, studies show that six or more
years of schooling for either parent is needed to lead to a substantial
reduction in child mortality (Martin et al., 1983:425).
Some studies, however, show the importance of the mother’s schooling
rather than the father’s.
Farah and Preston (1982:378) observed in the
Sudan that additional years of schooling of the father had a smaller effect
on the health status of the child than additional years of schooling of the
mother.
Similarly, it has been observed that the level of child morbidity
is inversely related to the educational level of the parent, especially of
the mother.
In Nigeria, the risk of contracting diseases of hygienic and
nutritional origin was three times higher among children whose mothers were
illiterate or had primary education than among those whose mothers had
secondary education (Chojnacka and Adegbola, 1984:805-806).
The
explanation given was that education probably brings improvements in
4
hygienic conditions and changes in those traditional customs which are
argued that education makes
unhealthy practices.
(Caldwell
------ (1979) also
-mother more capable of manipulating the modern world, that is, knowing
where the facilities are and becoming more likely to use them.
Thus,
In rural Ethiopia, women with formal education are very few.
this section focusses on the relationship between child morbidity and
literacy status of their mothers, as presented in Table 5.
From this
table, it can be observed that children whose mothers are literate have
The
higher reported morbidity than children ot illiterate mothers.
• ■ is
’ seen among males.
’ .
The same observations, that
largest differential
morbidity is higher among children of mothers with some schooling than
those whose mothers have no schooling, were made by Ramlah (1986) and
Bhuiya et...al. (1986).
However, regarding access to sanitary facilities,
Table 5:
Percentage of Sick Children by Sex and
Literacy of Mother
Female
Litexacy,status
Literate
Illiterate
27.8 ( 432)
28.9 (5420)
32.6 ( 399)
28.7 (5554)
30.1 ( 831)
28.8 (10975)
rather better conditions exist among children of literate mothers than
among those of illiterate mothers.
For example, 10 per cent of the
children of literate mothers have access to a latrine compared to 4 per
cent of children of illiterate mothers.
Therefore, sanitary conditions
alone would not be responsible for the observed variations between children
of literate and illiterate mothers.
Johansson and Mosk (1987) asserted
that the degree of reduction in morbidity is high at a public level where
most diseases depend on community sanitation conditions.
Thus, literacy
could only have a limited impact on child morbidity, since its effect is
seen at household level.
Hence, these differentials could be due to
under-reporting of diseases by illiterate mothers either because of
differences in perception of illness or because of their beliefs about
disease reporting.
Caldwell (1983) observed failure to report disease in
traditional Indian society, due to a belief that reporting diseases that
are categorized under ’supernatural power’ is not good or safe.
The types of disease present among the different groups of children
according to their mother’s literacy status is shown in Table 6.
It
appears that diarrhoea and gastro-enteritis are more prevalent among
children of illiterate mothers than those of literate mothers.
Among
children of young mothers (<25 years of age), most of the diseases are
found to be more prevalent among the children whose mothers are illiterate.
However, skin infection is reported more commonly among literate (younger)
mothers, and this may be due to different perceptions.
Malnutrition is
generally more prevalent among those whose mothers are illiterate (about 6
per thousand) than those whose mothers are literate (about 4 per thousand).
This supports the argument that education could bring a change in feeding
practices and nutritional level (Smith et al., 1983:104-106), though
economic status is probably higher for the literate mothers.
It also
seems to agree with studies in Colombia, India and Jordan, that showed a
positive relationship between female literacy and nutritional status
(Cochrane et al., 1980).
However, malaria seems to be prevalent among
5
children of literate mothers consistently over all age groups, while the
rest of the diseases do not show a clear pattern.
Table 6:
Prevalence Rate (per thousand) of Diseases Among Children
by Age and Literacy of Mother
Age.Gro_up,.jof....Mother
_____ 25-2.9_______ ______ 30.+.
Lit
i t-I i t ... ._ Lit.... Illit
Types of Disease
__ Lit.
Diarrhoea &
Gastro-Enteritis
Parasitic
Whooping Cough
Pneumonia
Other Respiratory
Eye Infection
Skin Infection
Malnutrition
Malaria
Measles
Other Diseases
Number of Cases
78.7
9.8
6.6
0.0
42.6
29.5
42.6
3.3
16.4
6.6
65.6
(305)
54.3
27.1
31.7
4.5
76.9
36.2
45.2
0.0
4.5
0.0
49.8
(221)
87.4
21.0
13.1
3.5
59.0
31.5
27.1
5.2
2.6
7.4
52.9
(2289)
79.1
20.4
13.6
3.2
55.9
19.7
24.7
6.1
3.6
6.4
56.9
(2793)
65.8
26.3
3.3
0.0
62.5
19.7
23.0
6.6
6.6
6.6
52.6
(304)
71.3
21.1
9.9
2.9
55.9
21.0
28.7
6.5
4.1
6.1
47.4
(5863)
Maternal Health Status
Table 7 shows the relationship between mothers' health status and
children’s health status.
The correlation is substantial.
It appears
that more than 40 per cent of children whose mothers were sick were also
sick themselves, but only 21 per cent of children whose mothers were not
This applies to
sick, or about half of the above group, were sick,
association between children’s
children of mothers of all ages.
rThis
.—-------Table 7:
Percentage of Sick Children by Mother’s
Health Status and Age
Age Group
of.Mother.
_____Maternal_.HeM^Jtatys-------
<25
25-29
30+
47.0 ( 827)
44.0 (1046)
43.0 (4372)
IllNPX.J11_
23.4 (1767)
21.5 (1969)
20.5 (7447)
One reason could be
health and maternal health could be for two reasons,
Since this study focusses
exposure to the same environmental conditions.
on children who are living with their mothers, it is obvious that they
would have access to the same facilities, such as water supply and
environmental sanitation.
The other reason is that there might be less
care and attention given to the child while the mother is sick.
It is
more likely that a sick mother has less time to spend with her child; this
could affect the child’s health.
Personal contact of the mother with the
In the case of
child may also be another reason for the observed finding.
6
a sick mother, the child has a higher risk of contracting the diseases due
to constant personal contact while feeding, cleaning or breastfeeding
particularly infants.
Maternal Characteristics and_Child.'.s Use .of Health..Facilities
The differential use of various treatment facilities depends on a
number of socio-economic factors, including cultural beliefs of the people,
access to the facilities and knowledge and attitude towards modern
medicine.
Maternal Literacy Status
Several studies assert that there is an association between maternal
education and a child's use of medical facilities. Osborn and Ameen
(1975) observed in urban Pakistan that education and income are important
determinants of use of maternal-child health facilities.
Caldwell (1983)
observed that education helps mothers to bring their children to modern
From
health facilities even before practising other traditional medicine,
their study in Tamil Nadu, India, Rao and Richard (1984) suggest that
education helps people to seek the right source of medical services.
Due to the small cell sizes when variables are cross-tabulated, the
relationship between the use of treatment facilities and maternal
characteristics according to common types of disease is difficult to
analyse in the present study.' Thus, this section only focusses on the use
of treatment facilities for the total morbidity experienced by the
children.
Among all sick children, about 37 per cent whose mothers were
literate received treatment from one of the facilities, compared to 33 per
This is
cent of those whose mothers were illiterate (Table not shown),
obtained from a direct question to those who reported being ill: ’Did you
receive treatment during the last 14 days?’
Table 8 shows the percentage distribution of all children who
received treatment, by different types of facilities in relation to their
maternal literacy status, while age of mothers is controlled.
Table 8:
Percentage of Sick Children Who Received Treatment by
Type of Facility and Literacy Status of Mother
_ _ 25-29______
Lit
Illit.„„
Treatment
—1.^7.^?....
Medical facilities
Pharmacy
Traditional Healer
Lay Treatment
Self Treatment
Unqualified person
Number of Cases
.Lit.. .. ..I.1.1.1..T'....
52.3
13.6
25.0
9.1
0.0
2.3
(44)
57.1
14.3
14.3
3.6
3.6
7.1
(28)
38.1
18.3
21.6
18.0
4.3
1.8
(278)
40.5
16.8
20.3
16.2
3.8
4.8
(291)
____ 3Q+____
„„Lit
Illit
40.0
15.0
25.0
5.0
10.0
5.0
(20)
39.7
13.0
17.9
18.5
7.1
5.9
(47)
This shows a marked difference in usage of medical facilities between
children of literate and illiterate mothers.
Children of literate mothers
have high usage of medical facilities.
This is probably because literate
mothers are more aware and have better knowledge of modern medicine than
7
illiterate mothers.
In contrast, among children whose mothers belong to
the old age group, there was little difference in use of medical facilities
between those of literate and illiterate mothers.
Thus the effect of
literacy on utilization of medical facilities is stronger for children of
As expected, lay and self treatment
younger mothers than older mothers.
An interesting
are more used by children whose mothers are illiterate.
pattern is that traditional medicine is still more preferred by literate
mothers who are less than 25 and above 29 years of age than by illiterate
mothers.
Table 9:
Percentage Distribution of Children Who Received
Treatment From Other Than Medical Facilities by
Reason and Literacy Status of Mother
Reason
Modern Medicine
Too Expensive
Disease Will Stop
By Itself
Used Traditional
Medicine
Too Far
Not Believe it Helps
Long Waiting Time
Thought Couldn't Get the
Required Medicine
Other
Number of Cases
Literate
Illiterate
_Total
20.9
26.8
26.4
20.9
24.5
24.2
14.0
14.0
4.7
0.0
12.6
11.8
8.2
4.1
12.7
12.0
8.0
3.8
2.3
23.3
(43)
3.9
8.0
(609)
3.8
9.0
(652)
For children who received treatment from other than medical
facilities, economic reasons were the most common reason, as shown in
Table 9.
It appears that for children whose mothers are literate, ’other'
is the most frequent reason, reported followed by 'too expensive' and
'disease will stop by itself.
However, the latter two reasons are
reported more frequently for children of illiterate mothers than literate
mothers.
Unlike findings in other developing countries, distance or
access to health services does not seem a typical problem, either for
literate or illiterate mothers.
However, other cultural and economic
barriers seem to be responsible for them not seeking medical facilities.
In summary, there was little differential according to sex of the
child for either morbidity of utilization of health services for sick
children.
The level of seeking treatment at health services was
relatively low.
Among parents who sought out one of the treatment
facilities for their sick children, more than half of them received
treatment from other than modern medical facilities. Furthermore, for
common diseases, infants and toddlers (aged 1 to 2 years) were more likely
to be taken to modern medical facilities than older children.
Although the children of literate mothers were more likely to be
reported as ill, these sick children (of literate mothers) were more likely
to be taken to modern medical facilities, than the sick children of
illiterate mothers.
The major single reason for sick children not to be
8
•• s.
taken for treatment to modern medical facilities was economic, that it is
'too expensive'.
The policy implications from the findings described in the previous
Research Note (Yohannes and Streatfield, 1988) emphasized health education
and improvements in sanitation.
The implications from the findings
presented here are that the Government's Primary Health Care programme,
while concentrating on expanding health services to the rural areas, needs
also to consider ensuring that the cost of treatment and medicines are
within reach of the population, otherwise the newly developed modern health
services may remain underutilized
REFERENCES
Bhuiya, A., S. Zimicki and S. D’Souza, 1986, "Socioeconomic Differentials
in Child Nutrition and Morbidity in a Rural Area of Bangladesh",
Jpurnal.„pf... Pediatrics, 32:17-23.
Caldwell, J.C., 1979, "Education as a Factor in Mortality Decline: An
Examination of Nigerian Data", Population Studies, 33(3) .-395-413.
, p.H. Reddy and P. Caldwell, 1983, "The Social Component of
Mortality Decline: An Investigation in South India Employing
Alternative Methodology", PopulaUon^^
: 185-205.
Chojnacka, :H. and 0. Adegbola, 1984, "The Determinants of Infant and Child
Morbidity in Lagos, Nigeria", SpclalSlence anOedicine, 19(8):799810.
Cochrane, S.H., D.J. O'Hara and J. Leslie, 1980, "The Effects of Education
on Health", WorldLBankStaff.JWording...Paper, No.405. Washington DC:
World Bank
Farah, A.Z. and S.H. Preston, 1982, "Child Mortality Differentials in
Sudan", Population and Develpa^^
8; 365-383.
Johansson, S.R. and C. Mosk, 1987, "Exposure, Resistance and Life
Expectancy.- Disease and Death during the Economic Development of
Japan, 1900-1960", Populations
41:207-235.
Kune, J.B., 1980, "Some Factors Influencing the Mortality of Under-Fives in
a Rural Area of Kenya: a Multivariate Analysis", Jpu™^^
Ped.iatrics, 26:114-166.
Martin, L.G., J. Trussel, F.R. Saluail and N.M. Shah, 1983, "Co-variates of
Child Mortality in the Philippines, Indonesia and Pakistan: An
Analysis, Based on Hazard Models”, PopulaXi™^^
37(3) .-417-432.
’ ”. and L.C. Chen, 1984, "An Analytical Framework for the Study of
Mosley, ■W.H.
Child Survival in Developing Countries", Population and Development
Review, Vol.10 Supplement:25-45.
Osborn, R.W. and S. Ameen, 1975, "Demographic and Socio-Economic Variations
in Health and Family Planning Behaviour in Urban Pakistan", Social
Science and...Mediclne 9(11/12) .-659-663.
9
Ramlah, H.M., 1986, Education and Use of .Healt h_.Seryices_.in.....Malays.i , M.A.
Thesis. Canberra: The Australian National University.
Rao, P.S.S. and J. Richard, 1984, "Socio-economic and Demographic
Correlates of Medical Care and Health Practices", Jpur.nal._of
Biosocial.Science, 16(3):343-355.
Smith, M.F., S.K. Paulsen, W. Fougere and S.J. Ritchey, 1983,
"Socioeconomic, Education and Health Factors Influencing Growth of
Rural Haitian Children", Ecology Q.LFpod..^^
13:99-108.
Waldron, I., 1983, "The Role of Genetic and Biological Factors in Sex
Differences in Mortality", in Sex Djfferent^ials...in Mortality, eds.
A.D. Lopez and L.T. Ruzicka, pp.141-164. Canberra: ANU.
Yohannes, A.G. and K. Streatfield, 1988, "Child Morbidity Patterns in
Ethiopia, 1983”, Child Survival Research Note No.lSCS, 12pp.
Canberra: ANU, International Population Dynamics Program, Dept, of
Demography.
■
.3
• rt-
10
Research Note on
CHILD
SURVIVAL
Number
20CS
Date
28 August 1988
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
ADAPTING THE SAFE MOTHERHOOD INITIATIVE
TO INDONESIAN SOCIETY
Terence H. Hull
Department of Political and fSocial Change,
Research School of Pacific
-c Studies,
The Australian National University
Note :
u (
A/VO
CEA/TPE
Child Survival Research Notes are brief discussions of
issues of current relevance to researchers and policy
makers concerned with problems of high infant and child
mortality m the world. The International Population
-^Dynamics Program, Department of Demography, The Australian
National University, distributes these notes with their
BlbllograPhic Circular. Production of the
, - J-d Survival Research Notes is made possible through
-ga^rant from the Ford Foundation (840-0893).
JJesponsibility for the content of Child Survival
Research Notes rests with the author(s) alone and not
the above-listed organisations.
The.Problem
”If it's not broken, don’t fix it.” Certainly words to
work by, but also a warning that it is easy to fall into the
trap of over-zealous criticism, especially when considering
new development initiatives.
The problem posed in this paper
is whether the recent Safe Motherhood initiative to reduce
maternal mortality is basically sound, or alternatively, is
misguided enough to require ’’fixing”.
§afe.._Motherhoc»d has emerged in the last two years as a
major public health initiative, and we now see workshops,
symposia, conferences, and project proposals demanding the
attention of policy-makers, donors and researchers throughout
the world. Safe Motherhood became the theme for a conference
in Nairobi in February, 1987, drawing sponsorship from a wide
range of international agencies.
In March, 1987 the then
Director-General of the World Health organization, Halfdan
Mahler, published an article in Lancet, summarizing the
recommendations of the conference.
Titled ’’The Safe
Motherhood Initiative: A Call to Action” it pointed to the
estimated half million deaths occurring to women in pregnancy
and childbirth each year, and called for immediate and
concerted international responses to meet this appalling
challenge.
A further conference in October, 1987, again in
Nairobi accepted this call, but broadened the mandate to
’’Better Health for Women and Children Through Family
Planning”. (Population Council, 1987).
In Indonesia a major meeting was held in June, 1988 to
consider the Safe Motherhood theme, and a number of proposals
are on the drawing boards to promote Safe Motherhood through
the Family Planning Program, the Department of Health, and a
variety of Non-Governmental Organizations.
It would seem
that this is an idea whose time has finally come.
The human
costs of maternal and early infant mortality are high1, the
priority of action to reduce them is clear. Superficially,
the recent initiatives have been widely welcomed by
development agencies and donors, and thus there would appear
to be nothing wrong, and nothing to ’’fix”.
In the next few pages I want to argue that appearances
may be deceiving, and the ’’Safe Motherhood” initiatives being
pursued with such vigour in Indonesia may be on the wrong
track.
This argument says nothing to challenge the data on
maternal mortality assembled to date, nor does it seriously
question the analyses of maternal and infant morbidity and
mortality presented in various conference documents.
Instead
it focusses on two related problems.
First, is the tendency
of the "Safe Motherhood” interventions to be seen in terms of
explicit top-down delivery of services from trained
professionals to recipient mothers, with a heavy emphasis on
clinic oriented, though community based, ante-natal care.
Dr Haryono Suyono has estimated that between 25 and 30
1.
thousand maternal deaths occur in Indonesia each year, or
about six percent of the world total.
1
Second, there is little attempt to understand the traditions
and meanings of childbirth, and in particular, no
acknowledgement of the roles and responsibilities of fathers,
families, and friends in the protection of mothers and the
newborn.
This paper argues that there is something
fundamentally wrong with much discussion of ’’Safe Motherhood”,
which needs to be fixed quickly.
.^.0.*^
.I.A.^.-O.g.^.
It is important at the outset to clarify the relevance of
tradition in the preparation of development projects. As
policy-makers in Indonesia sometimes contend "we do not want
to be an ethnic museum.” The purpose of development, in their
eyes, is to change and to do so in ways which reinforce
national identity, unity and resilience. Often this is taken
to be identical with the abandonment of local traditions in
favour of national models of modernization.
The problem comes
in identifying appropriate models for the construction of
modern ways.
If explicitly ’’Western” models are adopted,
there is a problem of adaptation, and the conflict with
nationalistic ideals.
It is partly for this reason that
Indonesia has rejected western notions of liberal democracy in
favour of an indigenous amalgam of ’’traditional” principles
known as Panca Sila democracy. The role of tradition in this
case is not to provide an unchanging fabric of social and
cultural relations to be protected against change (as
conservatives in some parts of the world espouse) but rather
to promote fundamental elements of traditional culture which,
when combined with new technologies and social structures,
will produce innovative modern institutions.
The task of matching appropriate traditional patterns to
new technologies is often difficult, and is certainly
controversial.
It requires a sensitive understanding of
traditional culture, as well as an evaluation of the minimum
requirements for adaptation of new technologies. In Indonesia
today the packet soup spices for Rajwon and Soto Madura
represent a successful form of adaptation where time-tested
recipes have combined with new methods of preservation to
provide a food product well adapted to the new time schedules
of urban families.
In contrast many people would question the
success of trying to blend traditional and modern elements in
lavatories, some urban architecture, wedding receptions, and
the youth culture.
Iradi.tional. Chi.l.dbeari.ng..:_ Variety across Cultures
’’Tradition” is not homogeneous across Indonesia, nor even
within the most populous island of Java.
Different regions,
social classes and ethnic groups have quite different ways of
organizing childbearing.
As such it is not possible to speak
of a single set of behaviours as ’’the” Indonesian or Javantradition, rather there are a wide range of practices.
In Yogyakarta, the heartland of Javanese culture, the
ante-natal period and parturition are times of concern not
only to the woman, but also to her husband, close relatives
and neighbours and birth attendent.
The community at large is
enlisted in ceremonies to appease spirits, appeal for God’s
protection, and protect the mother and child. As
Koentjaraningrat points out "In addition to being a religious
event, the Javanese childbirth is also a social event.”
(1985:104)
Throughout the Javanese cultural region ceremonies are
held around the seventh month of pregnancy, attended by the
traditional midwife (eg. Geertz, 1961: 87). These may be
extremely elaborate, in the case of urban elite (priyayi)
families (eg. Bratawidjaja 1988: 21 ft.), more modest among
relatively well-off villagers (C. Geertz, I960: 38-45) or very
simple affairs among poor villagers (V.Hull, in Appendix).
In
this ceremony (mitpni, tingkeban. or keba) the wife and husband
carry out a series of symbolic actions to ensure a trouble
free delivery. Koentjaraningrat stresses the ambivalence
reflected in the ceremony:
”On the one hand, it is a happy
anticipatory announcement of a birth, but on the other hand it
includes elements which stress the dangers of childbirth”
(1985: 103). Tradition dictates that both the wife and husband
observe taboos to protect the family.
The Javanese traditional birth attendant (dukun bayi)
combines skills in magic, massage and herbal treatment to
assist in the birth. Hildred Geertz describes a birth in East
Java as follows:
She spreads a mat on the floor or on a wide bedsized bench and seats the wife on it, and the
husband sits behind the wife upon a little bench
about six inches high, supporting her between his
legs as she leans back, straining to ’push’ her baby
out. The midwife meanwhile is saying the proper
magical spells for the protection of the parturient
woman and the baby and is firmly massaging the
woman’s legs, thighs, and abdomen.
The husband
takes a mouthful of certain special herbs . . . and
chews them into a paste; while his wife is in labour
he spits them onto her fontanel, giving her added
magical protection. . .
After the baby is born, the husband has the final
task, which only he may perform. . . He must take
the mat on which his wife has been lying (kppphan)
to a river to wash off the blood. There he must
burn some incense and some dry rice stalks, put a
2. The major Javan ethnic groups are the Javanese of Central
and East Java, the Sundanese of West Java and the Madurese of
the island of Madura and coastal regions of mainland East
Java.
3
flower and lime mixture (sekar fepreh) nearby, and
say a special spell taught him by the midwife to the
spirit (baureks.a) of the river.
There begin five days and nights of continual
entertaining, often with all-night card games, . . .
climaxed by the big pasaran ritual meal and
entertainment on the fifth day. For many women this
means severe strain and sickness, and occasionally
it can be a cause of maternal death. (1961: 87-88)
The descriptions of births in Yogyakarta presented in
Jaspan and Hill (1987: 10-12, 55-59) and Hull (Appendix) are
similar in most essentials, but contain some important
differences of terminology, ceremonies, and the descriptions
of the roles of husbands.
The five days immediately post-partum, and the series of
"birthday" ceremonies held in 35 day cycles thereafter are
traditionally intended to denote the envelopment of the new
child into the community for protection and as an expression
of community solidarity and harmony.
The husband has a key
role in all of these stages.
In Yogyakarta the husband takes
part in the major ceremonies, acts as host for the selamatan
ritual meals and all-night jag.ongan, and often assists in the
delivery. He will also immediately bury the placenta and
umbilical cord (the adik ari-ari, or symbolic younger sibling)
in the house compound, in an urn containing a variety of
materials of symbolic significance: a steel needle, pencil,
coin, sugar, rice and salt.
In my experience, husbands have been concerned to ensure
that their wives do not over-exert themselves, and have relied
on female relatives and neighbours to prepare snacks and
meals, and help with the baby, though as Geertz indicates it
is not unusual to find the new mother working in the kitchen
at some time during the five days post partum. In a
traditional sense, the continuous presence of neighbours and
kin "is felt to be an important part of the security with
which a newborn child is surrounded," (Jay, 1969: 31) and can
be seen as a response to recognised dangers surrounding
childbirth.
In other areas of Java the practices surrounding
childbirth are different.
In Madurese regions the traditional
midwife is called a "dukon rembi" (Niehof, 1985: 249-254) and,
according to Jordaan (1985: 81):
The husband’s assistance in childbirth is minimal.
He usually waits outside the house until the child
is born, after which he will whisper the addan and
Limmat [prayers] into both the baby’s ears.
If the
husband is present during the delivery itself this
has a special reason: to support his wife when there
are unexpected complications or to bite through the
umbilical cord (instead of having it cut by knife by
the midwife), to enhance the strength of the newborn
baby.
The placenta is buried by the father, in an earthen pot on the
top of which a small lamp is burned for seven or forty days
(Neihof, 1985: 239).
The mother is required to stay near the
house, which is protected by signs and spells laid out by the
dukon rembi.
A ceremony on the seventh day after the birth
is the occasion for naming the baby, and on the fortieth day
the dukon rembi hands over full responsibility for the care of
the baby to the mother. At this time restrictions on the
mother’s contact with outsiders are lifted, the postpartum
taboo against sexual intercourse lapses, and the baby is
formally introduced to neighbours, who join the family in a
ceremonial meal (rasol).
While the Madurese father is
apparently not so actively involved in the delivery as his
Javanese counterpart, his role in various ceremonies and
supportive activities is nonetheless clear.
While Javan people traditionally chose to give birth at
home, where spiritual protection was offered, the Ma’anyans of
southern Kalimantan sometimes give birth in small huts away
from the main village, to avoid putting their house into a
state of ritual pollution. Nonetheless, Hudson reports that
"The baby’s mother is attended by her mother or mother-in-law
. . . a midwife, and her husband,” (1972: 113). The midwife
is usually an old woman, but occasionally men serve in the
role.
In common with the Javanese, the Ma’anyan hold
ceremonies in the late stage of pregnancy for an easy
delivery, and parturition is regarded with great apprehension.
”A successful delivery is greeted with much relief by the
whole family.” (Ibid.)
In the 1930s Covarrubias found the birth customs of the
Balinese to be remarkably open, with the topic being freely
discussed in the community. (1937.- 124). ’’Frequently even the
assistance of a midwife is dispensed with and only expert
women relatives aid the woman. As a rule the husband should
be present.” The birth takes place with the woman leaning
against the midwife or a woman relative, and the placenta is
placed in a yellow coconut to be buried by the husband outside
the family’s sleeping quarters. A period of ritual impurity
follows for 42 days during which family members are enjoined
to be particularly careful, and make special offerings. At
the end of the period the mother and child are blessed by the
priest in an elaborate ceremony of purification.
Q..X0.
There are many elements of traditional childbearing
practices in Indonesia which are relevant to modern needs.
First is the common pattern of responsibility of the father
and traditional midwife for ceremonies in the ante-natal and
immediate post-natal period. These ceremonies are clear
expressions of concern for the safety of the mother and child.
Modern practitioners could encourage such practices, since
they are generally benign in their impact on the mother and
child, and might well build ante-natal and post-natal care
into the timetable of the ceremonies.
5
Second, among a majority of Indonesians, the father and
other close family members are often present at the birth,
giving support to the mother, and assistance to the midwife.
The family and the community have thus traditionally invested
time and resources in obstetric care -- albeit directed to
presumed spiritual sources of danger -- and might equally be
expected to organize resources for the care of women
identified by modern diagnostics as being at "risk”.
Third, tradition tended to unite people of different age
groups, social statuses, and backgrounds in common concern and
activity for the welfare of mothers and children. Modern
facilities tend to exclude people on grounds of lack of skill,
notions of hygiene, or for reasons of administrative
efficiency. By confining childbirth to ’’institutional”
spaces, and limiting access to "professional attendants” the
medical facilities often isolate mothers at precisely the time
when they need most support and attention.
Traditional birth attendants in Indonesia have long been
the object of ’’training” to teach them rudimentary principles
of hygiene and provide them with a basic kit for assisting in
deliveries. The first attempts at such training in Indonesia
were in 1807 (Riyadi, 1981: 37) This has generally been seen
as an interim measure to reduce major problems of infection,
to be eventually superceded by full-blown medical intervention
by professionals once resources are available. Traditional
attendants have not been integrated into the health care
system as acknowledged members of the health service team, nor
have their practices and beliefs been accepted as valid
elements of maternity care procedures in the local setting.
The gap between centrally planned health services and local
traditional care has always been large.
Finally, while the father is not always present at the
delivery, there is reason to believe that such a direct
involvement in the birth process might contribute to a form of
’’paternal bonding” beneficial to his relationship with both
his wife and baby.
The recommendations implicit in the above discussion
involve measures to locate maternal health services more
firmly in the home and community.
Specific responsibilities
and roles for the mother, the father, traditional healers, and
community leaders are the basis for such an approach.
These
are acknowledged as the primary responsibility points in
primary health care, and all professional staff from family
planning fieldworker to specialist practitioner are defined in
terms of their supportive and referral roles, (eg. see
Brownlea, 1983).
There are five major groups of
recommendations involving the involvement of family members,
the training of traditional attendants, the mobilization of
community resources to care for emergencies, the retraining of
professionals, and the development of adequate statistics on
problems of childbirth.
6
i • Involying Fathers and Other Family Members...... Where
feasible, fathers should be specifically involved in antenatal care, delivery, and post-partum care, in settings at or
Fathers should be educated in the risks of
close to home,
childbirth, and prepared to take timely action in case of need
for referral to a clinic or hospital. Sumapraja makes an even
stronger call: "Starting immediately we must accept the
husband as a valuable member of the obstetrical team, and stop
regarding him as someone intruding on the birth process"
(1988: 17).
When the husband is given a full understanding of
the physical and emotional changes occurring in the
childbearing process, he can become the most effective and
consistent source of support to his wife.
ii . Training Traditional Birth ,Attend.ants.„... Training of
traditional birth attendants should go beyond basic hygiene
and identification of risk factors, to integrate them in the
health services.
Benign traditional ceremonial practices
should be encouraged and combined with routine ante-natal and
post-natal care procedures.
This enhances the involvement of
the family, legitimizes the participation of traditional birth
attendants, and provides psychological reassurance to the
mother at a stressful time.
Education and procedures need to be directed to ensuring
that the quick referral of a patient in difficulties is
defined as a "system success" rather than the "failure" of the
traditional attendant. So long as the traditional attendant
remains marginal to the health care system, there will be a
tendency for hesitation in referral. The dukun today often
seeks to exhaust the options of the traditional practices
first, before turning in despair to the modern. By
integrating traditional practices with modern routines and
making the traditional attendant part of the health service
team, with formal roles at the POSYANDU service point, there
would be less loss of face in referring a patient.
iii. Empowering Local Leaders..,, In densely settled areas
well-served by publicly available transport services issues of
referral will depend on the ability of local people to
identify high risk cases, and the absence of prohibitive fees
in clinics and hospitals. Very poor families will need
adequate and accessible subsidies to ensure timely referrals
of women and infants in need. These might be provided via
some form of contributory insurance (Sumapraja, 1988: 19), or
through direct budgetary payment.
It should be the
responsibility of local leaders to ensure that such conditions
exist.
In more isolated communities local leaders must carry the
additional burden of ensuring that pregnant women identified
in the routine health screenings (through the integrated
health posts, or POSYANDU) or through difficulties in delivery
have adequate transport facilities available in case of
emergency. Community groups should also be organized to
provide blood for transfusions should that be necessary in
emergency care.
'7
One implication of this approach is that a high
proportion of births occurring in clinics and hospitals will
not necessarily be regarded as program success.
Rather, all
births should occur under supervision of a birthing team
including professionals, family members, and traditional
assistants as appropriate.
The place of birth will depend on
the needs of particular cases. Normal deliveries should occur
near or in the mother's home.
i v . Re train i..D.S
ng„„.Pr
o f es s i
Medically trained
.■E,jr..Q..£orici
_!. Staff.,...
jStci t f
professionals are currently taught not. to rely on families to
make critical decisions, not to collaborate closely with
traditional practitioners, and not to regard the family and
the community as having the primary responsibility for
maternal care. These lessons are inherent in their hospital
based residency training which stresses institutional care
under salaried professional responsibility. To change this
pattern will require a substantial program of retraining
involving the study and adaptation of traditional childbearing
procedures to achieve safer deliveries and better care.
Doctors and midwives used to giving instructions to large
numbers of women in a clinic setting will have to learn to
work with traditional practitioners and families in homes,
clinics and hospitals, according to the needs of particular
cases.
Included in the retraining of professional should be the
preparation and distribution of substantial manuals on
appropriate techniques and relevant issues in pregnancy and
birth. Some good models exist internationally (Parfitt, 1977;
Williams and Jelliffe, 1972; Hosken) but new manuals should be
based on the Indonesian national health system, and tailored
to local conditions in each province and significant ethnic
group.o
v. Statisti.cs and Research
A different approach to childbirth requires a different
approach to the collection of statistics.
The POSYANDU at the
local level should be made more efficient in registering
pregnancies and monitoring the progress of women through the
To some degree
ante-natal, delivery and post-natal stages,
this merely involves systematically recording the activities
of the dukun, bidan,
bidan, and health centre personnel.
3. The use of manuals to instruct in appropriate procedures in
childbearing has a long history in Indonesia.
Winklir (1890:
40) gave specific instructions on who might be allowed in a
delivery room: 1) the woman giving birth, 2) the traditional
midwife, 3) a servant, 4) the father, and 5) the woman’s
mother, if the woman requests her presence. Winklir, in
considering the problem of maternal mortality, also likened
child birth to a lottery where, out of 200 women, there would
be 199 winners, and 1 loser. (1890:35)
8
The focus of statistics should shift away from births in
institutions to monitoring of experience in identifying women
at risk, and recording outcomes.
Medical and social science research centres should be
encouraged to carry out studies of childbearing, Particular
attention should be paid to traditional practices, the
changing roles of traditional attendants, the health impact of
alternative practices, the attitudes and procedures of
medical professionals, and operations research on hierarchical
referral systems. Special studies will be needed to address
the problems of isolated regions (eg. Irian Jaya), scattered
islands (eg. NTT and NTB),
NTB) , and cases of resistence to
innovation.
In sum, the Safe Motherhood initiative is welcome as a
sign of government commitment to overcome the terrible
problems of maternal and infant mortality in Indonesia.
In
Rosenfield and Maine’s (1985) terms, the M is being returned
to MCH.
It should be seen as an opportunity to both
acknowledge the central role of families and communities in
preventive health care, and reform the government institutions
to support rather than supplant local efforts. This will
require a new understanding and adaptation of some long
established traditions of Indonesian society.
Bibliography
Bratawidjaja, Thomas Wiyasa. 1988. Upacara. Tradisiona1
Mas ya r aka. t. Jaw a..... (Traditional Ceremonies of Javanese
Society). Jakarta: Pustaka Sinar Harapan. Chapter 3
contains a detailed description of a t.i.ngkeban or mi ton i
ceremony, and Chapter 4 the tedak siten, with photos from
ceremonies carried out by well-to-do urban Javanese
f amilies.
Brown lea, Arthur. 1983. G r as s.r.op t s..._..I.n i. t .i.a t i ye s. in. Health.Care..:,
ban. and. .Rural. Models. th a t. Work...... Brisbane: Griffith
University.
Committee on Assessing Alternative Birth Settings. 1982.
Research. .Issues in the Assessment of .Birth Settings.,...
Washington: National Academy Press.
Covarrubias, Miguel. 193 7 (1974). Island. of .....Ba 11
Kuala Lumpur.- Oxford University Press.
Reprint.
Danforth, David N. and James R. Scott (eds.) 1986. Qbst.!e.tric„s
and Gynecplpgy... Fifth Edition with 86 authors.
Philadelphia: J. B. Lippincott.
Geertz, Hildred. 1961 . The Javanese. Family..: A. Study pf._..Ki.D.s.hip
and Soc.ialization,... Glencoe: The Free Press, pp. 85-92.
Geertz, Clifford. I960. The.Religion.of. Java. New York: The
Free Press. Chapter 4.
9
Hosken, Fran.
■ The„..Ch.ildbirth...Pictu^^
Hudson, A.B. 1972. Pad.Ju. E pat L... The _M a J a ny a n. of. Indonesian
Borneo.^. New York: Holt, Rinehart and Winston, pp. 113115.
Hull, Valerie J. 19/5. FertiXity...,....^^^
and.
the .Position. of Women in a_Jaya.nese Village.,. PhD Thesis
in Demography. The Australian National University. PP.
173-179.
Hull, Valerie J. 1979. ’’Women, Doctors, and Family Health
Care: Some Lessons from Rural Java. ’’ Studies. in. Family.
Planning.,... 10 (11-12): 315-325. See esp. pp 322-323.
Jaspan, Helen and Lewis Hill. 1987 . The Child in the .F.am.i.ly.: A
Study of....Childbirth and Child.-Rear ing. in. Rural. .Central
Hull:
Java in.„.the_..lat.e 1950s..,. Occasional Paper No. 14.
Centre for South-East Asian Studies. ISBN 0-85958-554-9.
Jay, Robert R. 1969. Javanese Villagers.! Social. Rel..a.t.i.ons. in
Rural Mod.i.okutQ..:... Cambridge, Mass.: The MIT Press. PP- 3031, 98-103, 218-223.
Jordaan, Roy E. 1985. .Folk.JJedicine in...M.adura (Indonesia).,. PhD
Thesis in Sociale Wetenschappen, Leiden University, pp.
80-87, 90-93, 97, 188-190.
Koentjaraningrat. 1985. Javanese.Cu 11ure.. Singapore: Oxford
University Press. pp. 100-108.
Mahler, Halfdan. 1987. "The Safe Motherhood Initiative: A Call
to Action". Lancet March 21, i: 668-670. Summary of
address to the Conference on Safe Motherhood, Nairobi,
Feb. 10-13, 1987.
Niehof, Anke. 1985. Women and. F e rt iJL ity.. in. Madura PhD Thesis
in Sociale Wetenschappen, Leiden University. Chapter 7
’’Pregnancy and Childbirth. pp. 221-254.
Parfitt, Rebecca Rowe. 197/. The Bir.th.„.P.r.i.mer_i A Source Book
ofTraditional ......andL..„.Alterna.ti.ye...JMethodsin... Laborand.
Delivery..,. Philadelphia: Running Press.
Population Council. 1987. Basic Documents .Better Health... for
New York:
Women. and. Children. thrpu.gJl..FaM-^
Population Council.
Note the advocacy document by Maggie
Black, and the Policy and Programme Implications by
Stephen Isaacs and Nuray Fineancioglu .
Riyadi, A.L. Slamet. 1981. Ilmu Kesehatan. Masyarakat..,.. (Public
Health). Surabaya: Usaha Nasional.
Rosenfield, Allan and Deborah Maine. 1985. ’’Maternal Mortality
- A Neglected Tragedy: Where is the M in MCH?” Lancet
July 13, ii: 83-85.
10
Sumapraja, Sudraji. 1988. Rawat. Ja.l.an_...K.eseha.t.a.n. Ibu
Professorial
Confirmation Speech, June 18, 1988, Faculty of Medicine,
University of Indonesia, Jakarta.
Suyono, Haryono. 1988. "Hamil dan Kehamilan dengan
Bertanggungjawab” (Responsibility in Conception and
Pregnancy). Prisma 17 (3): 57-62.
Werner, David and Bill Bower. 1982. Helping Health Workers
.
j
......a.^*..St?.
at the Village. Leyel,... Palo Alto: Hesperian Foundation.
i692} palo AltOj CA> 94302, USA.
See esp. Chapter 22.
Williams, Cicely and Derrick B. Jelliffe. 1972. Mother.. and
Ch,i.ld._.Heal.th...; .P.e 1 iy e r,yi ng. t he. Se ry i ce s..... L o nd o n .- Oxford
University Press. Chapter 10.
Winklir, F.C. 1890. Kitab.Iboe
Iboe. dan.Anak....
Anak...,. (Handbook of Mothers
and Children.) Soerabaija: Gebr. Gimberg and Co. Held at
the Perpustakaan Nasional, Jakarta, number xx, 183. See
esp. chapter 3.
11
Appendix. Extract from V.J. Hull,
1975:173-179.
Child Bearin.g...„.and....„Chil.d Rearing.. in Maguwohar jo^ Yogyakarta,..
19.72.-1973..^.
There are of course many aspects of having children that
are shared by upper and lower class people in Maguwohar jo.- the
joy and intimacy of the mother-infant bond, the pleasure of
playing with young children, the pride in having a child reach
certain levels of achievement. At the same time, in pregnancy
and childbirth, in care and education of growing children,
differences can be discerned in the attitudes and practices of
the rich versus the poor and the educated versus the
uneducated.
Pregnancy, and Chi 1 dbirth... Contrasting customs and beliefs
associated with childbirth can perhaps best be described by
the construction of two ideal types - a [poor] abangan woman
and a well-to-do priyayi (cf. Geertz, I960: Chapter 10). The
following descriptions are of course generalizations, but they
are based on the observation of several actual cases in
Maguwoharjo in 1972.
During her first pregnancy, the abangan woman will have
close contact with a dukun bayi (traditional midwife)
continually.
The dukun, an older woman who lives in the
village, will participate in the various pregnancy rituals,
It
the most important of which is held at the seventh month,
is said that these ceremonies help protect the fetus and to
one. Certain food
assure that the childbirth is an easy one.
taboos are observed during pregnancy** and
and,, more importantly, a
woman is very
verv careful co
to avoid saying harsh words to anyone or
harming anyone during this time, since it is believed this
would affect the fetus. A number of congenital deformities of
infants in Maguwoharjo were explained as arising from this.
The abangan woman would also take care to avoid places thought
to contain evil spirits, lest they enter her body and cause a
miscarriage. Other than this, however, little care is
exercised by the woman in terms of changing her life style,
and less so with each succeeding pregnancy. Pregnancy is not
thought of as being of particular danger physically, and
abangan. women continue to carry out their jobs right up until
the time the baby is due, even if it entails walking long
distances or carrying heavy loads.
When the time for the birth comes, the dukun is called to
the house, where she massages the woman's abdomen to get the
baby into the proper alignment. The actual birth is usually
attended by her husband and other members of her family as
well as the dukun. After the child is born, the woman is
again massaged to ensure that the placenta is expelled as
quickly as possible, since it is believed to be the spiritual
However, adherence to these taboos seemed to vary widely
even within social groups, and in fact individuals differed in
their opinions as 'to which foods were actually forbidden.
12
younger sibling of the infant (adik ari-ari). The umbilical
cord is cut only after the placenta has been expelled,® and
both the cord and the afterbirth are put in an urn with
traditional materials to be buried in the house compound by
the woman's husband.
There are a number of small rituals at specific intervals
following the birth, and the dukun visits the mother according
to these cycles -- usually coming every day to massage the
baby and mother' during the first 35 days, then massaging just
the baby up to the 70th day, then coming every 10 days to the
105th day after the birth.-7 The abangan mother would have
resumed her job after the first 35 days following the birth,
or possibly sooner. Most care in these matters is taken for
the first baby; subsequently the pattern may change.
The priyayi woman, particularly for her first baby, will
probably not enlist the services of a dukun. while she shares
many of the beliefs and taboos of pregnancy with her abangan
counterpart, her most important form of prenatal care will be
visits to the local clinic. Since it is unlikely that she
would have a job, most of her pregnancy is spent at home,
performing her usual domestic tasks. Her attitude toward
abangan birth practices in general is disdain! ulJ^ and when the
time for actual birth comes, she either goes to a Kl.inik
Bersalin (an obstetrics clinic) or has a bidan (trained
midwife, a paramedical position) come to her home.
The
equipment of the bidan is quite modern, but perhaps the
biggest difference in the actual childbirth process is a
social rather than a technological one. As Geertz (1961: 92)
pointed out, a birth supervised by a bidan does not involve
other family members. At one birth I attended, curtains were
drawn around the woman, and her husband was kept busy running
errands, bringing to mind the old image of the Western husband
rushing to "boil water" and "collect clean sheets". He did
not participate in the actual birth, and the woman's mother,
who lived in the next hamlet, was not in the house at all. . .
5. Traditionally a bamboo knife was used, though now many
dukun have a "kit” given them by the government during a
programme to upgrade traditional midwives.
6. Geertz (1961:89) writes that the dukun buries the placenta
with the recitation of a spell to control the spirit ot the
afterbirth.
In the cases observed in Maguwoharjo, however, it
was the husband who performed the burial.
7. This was the schedule followed by one dukun in Maguwoharjo
who explained it to me.
me. There are variations however, among
the various dukun.
8. Clifford Geertz (1960:359) cites this area of disagreement
as a major area of overt conflict between priyayi and abangan
beliefs.
In general, he claims, priyayi do not express open
disapproval of abangan customs, but many priyayi women's
groups have been very vocal in their condemnation of the
traditional, unhygienic practices of the abangan.
13
Priyayi women may also favour periodic massages after
childbirth -- massage in general is a trusted therapeutic in
Java, usually with good reason -- but she will probably not be
involved in the calendrical schedule which the dukun bayi of
the abangan woman followed. A priyayi husband, like the
abangan, does bury the placenta'3' and in Maguwoharjo births in
both groups were followed by a s.elametan and five nights of
visiting by neighbours -- men in the front room, the women in
the back with the mother and newborn infant.
9. This custom was also followed by highly educated urban
acquaintances in Yogyakarta, [in 1978, following the birth of
our son, we placed the placenta in a Javanese pot, with •? ,
traditional symbolic materials, and buried it on the shores of
the lake in Canberra where he was born, much to the amusement
of Australian friends, and to the pleased relief of many
Javanese friends.
14
Research Note on
CHILD
SURVIVAL
Number
21CS
Date
17 November 1988
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
DEMOGRAPHIC FACTORS RELATED TO LOW BIRTH WEIGHT
AMONG ABORIGINAL CHILDREN
Kim Streatfield
Child Survival Project,
International Population Dynamics Program,
Department of Demography,
The Australian National University
Ric Streatfield
Aboriginal Health Program,
Queensland State Health Department
Miranda Korzy
Child Survival Project,
International Population Dynamics Program,
Department of Demography,
The Australian National University
Note:
!Y
IN C
A/Vo
3 ■
Child Survival Research Notes are bried discussions of
issues of current relevance to researchers and policy
makers concerned with problems of high infant and child
mortality in the world. The International Population
Dynamics Program, Department of Demography, The Australian
National University, distributes these notes with their
regular Bibliographic Circular. Production of the
Child Survival Research Notets is made possible through
a grant from the Ford Foundation (840-0893)Responsibility for the content of Child Survival
Research Notes rests with the author(s) alone, and not
the above-listed organisations.
A .'
We wish to acknowledge the cooperation of the Aboriginal
Health Program, State Health Department of Queensland, through
which the data for this study were collected.
Introduction
It has long been established that infants born with
relatively low body weight are at increased risk of perinatal
and infant mortality, 'failure to thrive’, and ’slow learner
problems (Stanley and Hobbs, 1981:370,- Puffer and Serrano,
1973:41).
There has been considerable interest in the
question of whether or not such 'low birth weight’ children
(defined since 1976 by WHO as weighing less than 2,500 grams
at birth) can catch up with their heavier fellows. Data to
explore this question is rarely collected, but some such
studies suggest that they do not catch up but remain
permanently relatively undernourished, following low
projectile growth curves.
Even in well nourished societies, there are always some
low birth weight babies, due to a variety of factors including
prematurity (born before full term), genetic factors, etc.
In
Australia as a whole, the proportion of low birth weight
births is estimated to be about six percent, fairly typical of
developed countries.
This can be compared to about twenty
percent in developing countries as a group, though some of the
poorer such nations may have 40 to 50 percent (UNICEF,
1988:66).
The data used here are from eight Aboriginal communities
in Queensland.
These communities have participated in the
Aboriginal Health Program of the Queensland State Health
Department since it commenced in 1972.
The data refer to
births from the period 1978 to 1987. The proportion of the
births which were low birth weight was one in five overall,
ranging from one in eight to one in three among the various
communities.
The purpose of this paper is to explore the role
of certain demographic factors, maternal age, parity, and
other factors such as length of gestation, in these low birth
weight births, and in the variation across communities.
Trends,, in Birthweight
The data for the 1,551 births over the decade 1978 to
1987 show an average of 20.1 percent of births being defined
as low birth weight (LBW) (Table 1), with an average birth
weight of 2,932 grams.
The trend over time has been for a
fluctuation within a range of 17.0 percent to 23.9 percent
LBW, or 2,882 gm. to 3,011 gm.
These birthweights are quite comparable with the average
birthweights from a number of studies in Asia and Africa,
where average birthweights ranged from around 2,700 Gm. for
poor Tanzanians to 3247 for upper class Bombay Indians
(Ebrahim, 1983:36).
The above rates are somewhat lower than
average European populations, which tend to exceed 3,200 Gm
(ibid). The World Health Organization references, based on
mixed populations (racially and economically) in the U.S.A,
have a median birthweight of 3.2 Kg. for girls, and 3.3 Kg.
for boys (WHO, 1983:75,81).
1
The interpretation of birthweight data must be undertaken
in the light of the knowledge that birthweight is also
determined by the stature of the mother.
In general, ’’the
baby of a short woman is lighter and has less vitality and a
lower survival than that of a tall woman.” (Ebrahim,1983:40).
Table 1
Trends in Percent Low Birthweight Births, and in Average
Birthweight (and Standard Deviations)
Year
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
Percent
LBW
20.....1
17.8
19.3
23.9
22.8
18.1
21.7
20.7
17.0
19.2
20.1
Average
Birthweight
(Gm. )
2951
2982
2882
2903
3011
2832
2898
2949
2884
2917
Standard
Deviation
(Gm . ')
593
664
632
658
655
686
720
63 7
663
648
Number
1512
135
140
130
153
154
166
155
153
157
169
Mo.ther...’.s. Age.at. time....of ...child..! s..„. birth.
The reasons for the observed levels of low birth weight
infants are of considerable importance. Childbearing at an
early age is believed to be one such factor.
In fact it is
now generally accepted that the major risk of producing a low
birth weight infant occurs within two years of menarche
(Ebrahim, 1983:41).
The data in Table 2 indicate a clear positive relation
between mother’s age and birth weight with the infants of
mothers under age 20 being, on average, up to half a Kg. less
than those of mothers aged 35 and over.
In terms of percent
of births falling below a particular threshold, there is a
clear pattern of younger mothers being at much greater risk of
producing a low birthweight, and especially a very low
birthweight infant.
There is a well documented relation between mothers
weight (and height) and the birthweight of her children (Shah,
1981:53), and there is evidence that the weight of Aboriginal
women increases considerably with age (after puberty). A
study from the Kimberley region of West. Australia indicated
that Body-Mass Index (Kg./sq.m.) increased by some 50 percent
from 18.5 for 15-19 year olds to 27.6 for women aged 35 and
over (Rutishauser and Mackay, 1986:S9). This heavier maternal
2
weight may result in heavier infants.
The cause of the weight
increase may, however, be important.
There is some reason to
believe that the onset of diabetes among some relatively young
adult Aboriginal females may be leading to birth weight
increases which do not necessarily reflect healthier infants
(R.Streatfield, personal communication). It is hoped that
further analysis of this data set, and related data sets may
throw some light on this issue.
Table 2
Average Birthweights (& S.D.s), Percent Low Birthweight
(<2,500 Gm.), and Percent Very Low Birthweight (<2,000 Gm.),
According to Age of Mother
Mother’s Age
Average
Birthweight
Std .
Dev.
Percent
Low BW
Percent
Very Low BW
Up to 14
15-19
20-24
25-29
30-34
35-39
40 and over
2711
2766
2962
3077
3077
3143
3225
870
638
626
641
779
614
428
24.1
23.5
19.8
14.2
19.8
17.2
0.0
17.2
10.2
6.0
3.9
8.8
3.5
0.0
Table 3
Average Birthweights (& S.D.s), Percent Low Birthweight
(<2,500 Gm.), and Percent Very Low Birthweight (<2,000 Gm.),
According to Parity of Mother
Parity
0
1
2
3
4
5
6 or more
Average
Birthweight
Std.
Dev.
Percent
Low BW
Percent
Very Low BW
2790
2913
2947
3072
3044
3075
3268
659
625
672
616
735
673
604
23.1
20.6
19.9
13.8
14.1
22.0
6.4
9.7
6.6
7.3
4.9
6.4
4.9
2.1
3
Parity of Mother at time_oXJbirt_h
In this study parity was recorded to mean the number of
livebirths the mother had delivered before the index child.
As parity is closely related to age of mother, it might be
expected that there would be a positive relation between
average birthweight and parity. The data in Table 3 show such
a pattern. The positive relation between parity and
birthweight does not appear to be as strong as between
maternal age and birthweight, though first births (parity 0)
show greater levels of low birthweight than higher parities.
Gestation^
As mentioned above, there are two main reasons for low
birthweight. One is retarded foetal growth, the other is when
the infant is born early, prematurely, though it should be
noted that the term ’’premature” is also sometimes used to
refer to low birth weight infants born at full term.
’’They
There are a number of reasons for premature birth:
hard
include high maternal blood pressure, acute infections,
physical work, or multiple births....Premature infants are
usually thin, have muscle weakness, and a tendency to low body
temperature.
They may have difficulty suckling. The rates
for infections of various kinds, particularly in a poor
unhygienic environment, are considerably higher than for full
term babies, and death rates are high”. (Cameron and
Hofvander, 1983:3).
As a foetus grows rapidly during pregnancy, especially in
the latter stages (third trimester), it stands to reason that
being born even a few weeks before term may result in a weight
considerably below that which would have been reached by full
term.
The data in Table 4 show a rapid increase in average
birth weight according to length of gestation.
The average
weight is just 1.2 Kg. for the group of births below 32 weeks,
rising to 2.9 Kg. at 38 weeks and 3.2 Kg. at ^+0 weeks. This
pattern is roughly consistent with the observation that the
mamximum rate of foetal growth is during the 32-38 weeks of
pregnancy ’’when the weight virtually doubles” (Ebrahim,
1983:37).
The significance of the month between 34 and 38 weeks is
well illustrated by the decline in proportions of low
birthweight births from around four out of five at 34-35
weeks, to less than one in five at 38 weeks, and one in twenty
at 39-40 weeks.
Thus it is vital to ensure a gestation length
as close to 38-40 weeks as possible.
4
Table 4
Average Birthweight (& S.D.), Percent Low Birthweight
(<2,500 Gm.) and Very Low Birthweight (<2,000 Gm.),
According to Length of Gestation
(Weeks Since Last Menstrual Period).
Gestation
Average
Birthweight
Std .
Dev.
Percent
LBW
Percent
Very LBW
N
Up to 31
32
33
34
35
36
37
38
39
40
41
1202
1684
1986
2071
2268
2598
2831
2915
3007
3231
3285
661
384
555
<457
331
391
488
434
380
512
576
95.7
100.0
75.0
80.0
79.3
39.3
25.4
16.8
5.6
5.2
7.7
91.3
87.5
50.0
45.0
17.2
4.8
3.0
1.1
0.0
0.6
0.0
46
16
8
20
29
84
67
179
90
543
26
With length of gestation being so powerfully related to
birthweight, it might be hypothesized that the lower average
birthweights associated with young maternal age and low parity
might operate through higher rates of prematurity, i.e.,
shorter gestation periods.
This appears to be the case, as the average gestation
period for births to mothers of different ages increases from
37.5 weeks for those under 15 years, to 38.1 weeks (15-19
years), to 39.0 weeks for those 35 years and older. A similar
pattern applies to parity with the average gestation varying
from 38.2 weeks for parity 0, 38.2 to 38.6 weeks for parities
1-5, and 39.5 weeks for parity 6 or more. Analysis of
variance of these groups indicates differences significant at
the 5 percent level for both variables.
While the average gestation periods for the youngest
mothers and lowest parity births appear to be only slightly
shorter than for older mothers and higher parity births, those
few weeks can account for several hundred grams in
birthweight, and it is differences of this scale which
distinguish the young mothers and first births from the rest.
Thus it can be concluded that a substantial part of the link
between young maternal age, low parity births, and low
birthweight operates through such births resulting from
gestation periods somewhat shorter than average.
As mentioned above, a low birthweight infant may have
been ’born early’, or it may have not developed fully for its
gestational age. Such a small for gestational age (SGA) baby
(variously called a growth retarded, small-for-dates,
dysmature or malnourished foetus) is defined as such because
5
it becomes the newborn baby which weighs less than a
particular but arbitrary mass for gestational age.
The best
known standards of mass for gestational age are those given by
Lubchenco for her population of babies at Denver, Colorado.
In terms of these standards, babies are considered small for
gestational age if they are at or below the 10th percentile of
mass for gestational age (about 2,400 Gm. at 38 weeks, or
2,600 Gm. at 40 weeks) (Philpott, et al., 1978:46).
More recently, the Australian Commonwealth Department of
Health has developed a series of intra-uterine growth charts
which take account of various factors known to affect
birthweight.
These factors include sex of the infant,
maternal height, and parity of the birth, but they are not
presented separately by race (Comm. Dept, of Health, 1985).
The following charts (I-IV) illustrate that the infants
born to Aboriginal mothers in the present study, assuming they
are of medium height, fall between the 10th and the 50th
percentiles of mass for gestational age.
The question arises as to what is the relation between an
infant being born low birth weight and its various body
functions such as respiratory capacity, heart rate, etc.
The
most widely used method of assessing the physical condition of
a newborn is the scoring approach introduced by Apgar in the
1950s (Apgar 1953).
It is a score from 0 to lObased on adding
individual scores between 0 and 2 for each of the following
five criteria of heart rate, respiratory effort, muscle tone,
reflex irritability, and colour of the infant. The scores are
evaluated at 1 minute after delivery, then again at 5 minutes,
and sometimes 10 minutes.
Apgar scores of 7 to 10 indicate no depression of
function, 4 to 6 indicate moderate depression, and below 4
indicates severe depression (Romney et al, 1975:699).
There
has been some disagreement about the value and interpretation
of the Apgar score (see Gray, 1987: 11), but as the data are
available here they will be examined
The data in Table 5 shows a clear positive relation
between birth weight and Apgar scores at 1 and 5 minutes,
though the threshold appears to be for birth weights below
This pattern is similar to that noted by Gray
2,000 gm.
(1987:10), though the levels here are lower than in his study
area of N.S.W.
6
Table 5
Average Apgar Scores (& S.D.s) at One and Five Minutes, and
Percent Births Scoring Less Than 7 (’Moderate Depression')
at 1 Minute, According to Birthweight
Birthweight (gm.)
Up to 1499
1500-1999
2000-2499
2500-2999
3000-3499
3500-3999
4000 & over
...Apgar. 1. minute
% <7
Average (S.D.)
5.4
5.5
7.8
7.7
7.7
7.5
7.0
(2.5)
(2.7)
(2.0)
(2.0)
(2.1)
(2.0)
(2.3)
57.7
59.5
18.9
21.9
23.0
25.7
32.1
....... .. (d..3...*.
Average (S.D.)
7.0
7.6
9.2
9 . -4
9.3
9.3
9.
(2.7)
(2.1)
(1.3)
(1.1)
(1.2)
(1.4)
(0.9)
Table 6
Apgar Scores (Averages, and Percent below /)
According to Age of Mother, and Parity.
Apgar1.minute___
Average
Percent < 7
Apgar. 5. min.,.
Average
Mother’s.Age..:
Up to 14
15-19
20-24
25-29
30-34
35-39
40 & over
6.4
7.4
7.6
7.6
7.3
7.9
8.2
26.8
23.8
21.8
28.7
24.1
16.7
8.9
9.2
9.3
9.3
9.2
9.2
9.7
Parity.:.
0
1
2
3
4
5
6 or more
7.1
7.8
7.9
7.3
8.0
7.1
7.8
33.5
20.0
21.5
25.0
18.0
34.2
19.6
9.0
9.4
9.4
9.0
9.5
9.1
9.4
44.4
The figures in Table 6 show the expected pattern of lower
Apgar score for very young mothers (under age 15 years),
though mothers aged 15 to 19 years are about average,
The
first born infants also show a relatively low score, and
These
conversely have a high proportion of low scores.
7
patterns are consistent with patterns of low birthweights
which are known to be closely associated with Apgar score.
Community
The eight communities in this study vary considerably in
terms of distance from urban centres, as well as environmental
conditions.
The variation in average birth weight is greater
than half a Kg., while the proportions low birth weight range
from 13.0 to 37.8 percent (Table 7).
Table 7
Average Birthweights, Proportions Low Birth Weight Births,
Proportions LBW among 'Fullterm’ Births,
According to Community
Community
Average
Birthweight (S.D.)
% Very
LBW
% LBW
% LBW
(Fullterm)
Diff%
A
B
C
D
E
F
G
H
2943
2760
3025
2933
2839
3091
2459
2913
(643)
(567)
(610)
(694)
(584)
(611)
(755)
162.4.).
6.8
5.9
5.1
7.4
6.7
3.6
21.6
18.1
35.3
13.0
24.3
23.9
14.4
37.8
19., 4
7.9
14.8
4.9
11.6
7.7
4.9
5.9
.7.,. 9,
(10.2)
(20.5)
(8.1)
(12.7)
(16.2)
(9.5)
(31.9)
..(..11...?...5).
TOTAL
2932
(656)
6.8
20.1
7.5
(12.6)
Note:
'Fullterm’ here means 38 weeks or more gestation.
The pattern of percent LBW among fullterm births (7.5%
for all eight communities) shows that about two thirds (12.6%
of 20.1%) of the LBW births are due to being premature, This
factor is particularly important in communities B, E and G.
It was seen above that premature births are more likely
with first births, or births of young mothers, so the question
arises as to how important are these factors of low parity and
young maternal age in these high % LBW communities.
In Table
8 it can be seen that there is a wide range of proportions of
births occurring to mothers under age 20 years, but the
pattern is not consistent with the patterns of low birth
weight.
There is a narrow range of proportions of parity 0 or
1 births, but the pattern does fit rather better with the
pattern of LBW, where proportions of parity 0 or 1 births of
over 60 percent are related to high levels of LBW.
8
Table 8
Proportion of Births to Mothers Aged Under 15, Under 20;
Proportion of Births of Parity 0, 0 or 1,
According to Community
Community
Percent
under 15
Percent
under 20
Percent
Parity 0
Percent
Parity 0,1
Av.BthWt
@40 wks
A
B
C
D
E
F
G
H
2.4
0.0
0.9
6.3
6.8
1.2
0.0
1.8
34.2
27.1
33.9
34.6
43.7
28.6
36.7
3SL8
33.5
37.8
36.4
38.2
43.7
29.3
41.4
31.,.9
54.6
60.0
55.5
58.5
3248
3085
3322
3301
3065
3321
2967
3203
61 . '-i
51.6
65.5
57.1
Except possibly in community E, young maternal age does
not appear to be the cause of the high levels of LBW, but the
stronger association of LBW is with low parity births.
conclusion
These data from eight Aboriginal communities in
Queensland indicate a level of low birthweight fairly constant
at some three and a half times higher than the general
Australian population. Many of these low birthweight births
are associated with reduced physiological capacity as
indicated by Apgar scores.
Demographic factors associated with risk of low
birthweight are a maternal age at childbearing of less than
twenty years, and first order birth.
It appears that both of
these demographic factors are associated with low birthweight
via births being of shorter than normal gestation period.
While low birthweight can result from ’premature’
delivery, it can also result from less than adequate foetal
development, producing a ’small-for-gestational age’ baby.
The comparison of birthweight by gestational age of the study
children with the Australian intra-uterine growth charts,
indicate quite clearly that foetal development of these
Aboriginal children is up to several hundred grams below the
average (50th percentile) for Caucasian children at all stages
of gestation above 32 weeks.
The levels of low birthweight among the eight communities
vary considerably, and it is difficult to tease out the
precipitating factors in each case. While the communities
vary somewhat in the proportions of first births, and births
to young mothers, the patterns are not consistently linked to
the levels of low birthweight.
It does appear that there is
no single factor underlying the low birthweight births common
to all communities, but in some cases it is more a problem of
poor intra-uterine development, in other cases, more a problem
of premature delivery.
It is not suggested here that demographic variables are
the only factors of importance in the low birthweight story,
but there are patterns consistent with well documented
demographic patterns of infant mortality, though the parity
pattern does not fit.
Other factors, about which no data are available here,
include behaviours such as cigarette smoking and alcohol
consumption during pregnancy. Cigarette smoking by the
mothers in this study is known to be common during pregnancy,
and it has been observed elsewhere that the proportion of low
birthweight infants born to mothers who smoke more than 20
cigarettes per day is twice that of infants born to non
smoking mothers (Davies et al., 1976:385-387).
The incidence of low birthweight has also been observed
to increase by 2.7 times among infants of mothers who consume
excessive quantities of alcohol during pregnancy.
The risk of
low birthweight was increased by 1.8 times in mothers who only
smoked, whereas it was 3.9 times as great in mothers who both
smoked and consumed alcohol (Sokol, et al., 1980: 135-145).
At this stage, data on such behaviours are purely
anecdotal and are not available to be incorporated into the
analysis, but it is necessary to acknowledge their potential
importance in the problem of low birthweight births.
Future analysis on this data set will follow the
development of these low birthweight infants through the first
few years of life.
There are numerous questions remaining to
be answered about whether premature, but normal-for-dates,
infants are more likely to ’catch up’ to normal rates of
physical development than small-for-dates infants.
There are
also questions regarding the relationship between birthweight
of the mother, and the birthweights of her children, and how
low birthweight might better be predicted, and hopefully
prevented.
REFERENCES
Apgar, V. (1953), 'A proposal for a new method of evaluation
of the newborn infant’, Current Research in...Anaesthesia
and Analgesia , 32:260-267.
Cameron, M. and Y. Hofvander (1983) Manual on Feeding Infants,
and Young. Chi..ldren..i. Oxford Medical Publications, Oxford
University Press, Delhi, Nairobi.
Commonwealth Department of Health, (1985) I.nt_ra-uteri
Charts. Aust. Govt. Publ. Service.
10
Growth
Davies, D.P., O.P. Gray, P.C. Ellwood, and M. Avernethy
(1976), ’Cigarette smoking in pregnancy: associations
with maternal weight gain in foetal growth', The Lancet,
Vol.l: 385-387.
Ebrahim, G.J. (1983) Nu t r i.t ip.n _i n Mot her and Child Health.,
Macmillan Tropical Community Health Manuals, London and
Basingstoke.
Gray, A. (1987), ’Assessment of Risk for Newborn Aboriginal
Children’, Working Paper No.3, Aboriginal Family
Demography Study, Dept, of Demography, RSSS, Australian
National University, pp.36.
Philpott, R.H., K.E. Sapire, and J.H.M. Axton, 11978)
Obstetrics^ Ea.ni.ilX Planning and Paediatrics , University
of Natal Press, Pietermaritzburg.
Puffer, R.R. and C.V. Serrano (1973), Patterns of Mortality. in
Childhood,, Scientific Publication No.262, Pan American
Health Organization, Wash.
Rutishauser, I.H.E. and H. Mackay (1986), ’Anthropometric
status and body composition in Aboriginal women of the
Kimberley region', Med. J. Aust. Special Suppl., June 23:
S8-S10.
Shah, K. (1981) ’Maternal nutrition in deprived populations’,
Assignment Children 55/56 (2); 41-72.
Sokol, R.J., S.I. Miller, and C. Reed, (1980), ’Alcohol abuse
during pregnancy, an epidemiology study’, Alcoholism
Vol.4: 135-145.
World Health Organization, (1983) Measuring Change in
Nu t r i t i on a 1. Status., Geneva.
11
3080■
^ 2750-
2580-
■
(
(14 15-1? 20-24 25-23 38-34 35-33 48®
___________ ftge_____________
ss®
.
■
Ill
12
Chart 3: Average Birthweights According to Length of Gestation
3580 t
M-
I
t 2500 Average
(C"'’ 2NB-
d
1500-
1000
+
(31 32 33 34 35 36 37 38 39 48 41
________ Gestation________
h Chart 4: Average Apgar Score According to Birthweight
7,5?-
Apgar 6,5(1 Rin.)
.w
H
kS
5.55
>8
<1499 1580- 2B 2580- 3808- 3588- W
1999 2499 2999 3499 3999
Birthweight (Gm,)
13
-x.
Research Note on
CHILD
SURVIVAL
Number
22CS
Date
16 February 1989
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
MATERNAL MORTALITY IN INDONESIA:
A REVIEW OF RECENT EVIDENCE
Dr. L. Ratna Budiarso
PUSLIT Ekologi Kesehatan,
Badan Litbang Kesehatan,
Department of Health,
JI. Percetakan Negara No.29, JAKPUS,
Indonesia
Note:
s'
AND
CENTp£
/-’CaI.V ■ ‘L
Child Survival Research Notes are brief discussions of
issues of current relevance to researchers and policy
makers concerned with problems of high infant and child
mortality in the world. The International Population
Dynamics Program, Department of Demography, The Australian
Rational University, distributes these notes with their
Regular Bibliographic Circular. Production of the
flhild Survival Research Notes is made possible through
fa grant from the Ford Foundation.
Responsibility for
the content of Child Survival Research Notes rests with
the author(s) alone, and not the above-listed organ
isations .
INTRODUCTION
Each year worldwide 500,000 women of reproductive age die
as a result of pregnancy or childbirth, and almost all of
these deaths occur in developing countries. The maternal
death rate for women in developed countries is between 5 and
30 per 100,000 live births, while in poor countries the figure
reaches 50 to 800 per 100,000 (Mahler 1987). The risk of
death during the childbearing period can be expressed as
between 1 in 50 women and 1 in 14 women in developing
countries, compared to only 1 in 4000 to 10000 women in
industrialized countries.
The level of maternal mortality in Indonesia is not known
with precision because there no adequate system of vital
registration covering the whole nation. The limited data
available tend to come from various studies such as hospital
based records, health surveys, and population based
prospective studies to monitor maternal deaths. Each of these
approaches has its own strengths and weaknesses, and the
various methodologies are not strictly comparable, but in the
absence of more reliable figures, analysts are forced to use
these data, whatever their shortcomings, to gain a picture of
the level and causes of maternal mortality in Indonesia. This
paper provides a brief review of the state of the art.
DEFINITIONS
1. Maternal Mortality
Maternal mortality, according to the International
Classification of Diseases (Rev. 19/5: ICD IX), includes all
female deaths occurring during pregnancy, or within 42 days of
termination of pregnancy, irrespective of the duration and
site of the pregnancy, from any cause related to or aggrevated
by the pregnancy or its management but excluding accidental or
incidental causes.
In general, lay respondents to survey questions report as
maternal deaths only those caused directly by difficult
pregnancies or childbirth, but they are less likely to report
maternal deaths which appeared to have indirect obstetric
causes, even if, from a medical viewpoint, the apparent cause
arose from or was aggravated by the pregnancy. To obtain
accurate data on cases of maternal mortality cases, trained
medical interviewers are required to collect complete medical
histories.
2. Causes of Death
As indicated above, causes of maternal deaths are
differentiated into two groups:
a. Direct Obstetric Deaths, are those resulting from
obstetric complications of the pregnant state
(pregnancy, labour and puerperium), from
interventions, omissions, incorrect treatment, or from
1
a chain of events resulting from any of the above.
(ICD IX 640-646 and 651-676).
b. Indirect Obstetrical Deaths, are those resulting from
preexisting diseases or conditions, or diseases or
conditions arising during pregnancy, but not
physiologically resultant from the pregnancy (ICD IX
647-648).
Death certificates are supposed to include information on
all illnesses and conditions related to death, including
accidental or incidental conditions.
In analysing these
causes distinction is made between direct, intervening
antecedent, and underlying causes. Depending on the degree of
detail of information available and the capability of the
diagnostician, each cause can be coded according to the Basic
Tabulation List of the ICD IX, or more specifically in 3 or 4
digit code subcategories.
3. Maternal Mortality Measures
The commonly used Maternal Mortality Rate (MMR) as
described by the WHO covers both direct and indirect obstetric
deaths, expressed as a ratio to 1000 live births (ICD IX, Vol
1). Unfortunately, there is a great variety in the calculation
of the MMR by various analysts. The authors of a study in
Bandung (Alisyabana, et. al., 1983) include only direct
obstetric deaths compared with the total number of live
births. In the Houshold Health Surveys of 1980 and 1986 the
rate includes deaths associated with abortions (ICD 630-639),
and pregnancy, birth and neonatal complications (ICD 640-646,
651-676). Again, the rate is expressed in terms of live
births.
It is probably useful to draw the distinction between
the Maternal Mortality Rate and the Maternal Mortality Ratio
(MMO) (following Fortney, et. al. 1985):
MMO = {Annual Maternal Deaths/Annual Live Births}+1000
MMR = {Annual Maternal Deaths/Mid-year Population of
Women aged 15-44}*1000
The MMO reflects the hazards of pregnancy, while the MMR
relates deaths to the total number of women assumed to be at
risk of experiencing a maternal death.
If, as in the case of
Indonesia, fertility is falling rapidly, the MMO would tend to
decline at a slower rate for a given downward trend in
maternal deaths, than would the MMR.
REVIEW OF SOME ESTIMATES OF MATERNAL MORTALITY
In the past ten years a growing number of studies in
Indonesia have procduced estimates of the level of maternal
mortality. Among the most commonly cited are:
Type
Name
Year
Rate
Per
Teaching Hospitals (12)
Ujung Berung
Bali
Household Health Survey
Sukabumi
Household Health Survey
East Nusa Tenggara
Irian Java
1977/80
1978/80
1980/82
1980
1982/83
1985/86
1986
1986
3.7
1.7
7.2
1.5
4.7
4.5
11.4
6.6
1000 DEL Rec.
1000 LB Pro.
Pro.
Ret.
Pro.
Ret.
Pro.
Ret.
Note: Type: Rec.- Hospital Records, Pro.- Prospective.
Ret.- Retrospective.
DEL
LB
Deliveries
Live Births
Sources: See bibliography.
The studies cited above give a wide variety of estimate of
maternal mortality, depending on the data source used and the
technique of calculation. The hospital study is based on
women giving birth in hospitals, whether as normal delivers or
as emergency referrals. The Ujung Berung and Sukabumi studies
both used prospective data collection methods, but the
differences in results could be the result of different
systems of project administration and small sample sizes.
The maternal mortality ratios produced by the Household
Health Surveys of 1980 and 1986, based on one year recall of
deaths in the household, obviously could not avqid the problem
of under-reporting of events and mis-reporting of dates.
Nonetheless, the ratio for 1986 is much higher than that from
the survey six years earlier, giving rise to the question of
whether the actual level of maternal deaths has increased.
The answer is that the very different approaches to recording
diagnoses of death have produced a spurious apparent trend.
In 1980 only one cause of death was recorded for each case,
and in general this was the most immediate medical condition
........
,
In 1986, in addition to the immediate cause,
prior
to death.
all intervening conditions, and pre-existing problems were
also recorded, and through these records it became clear that
a large number of female deaths occurred during or soon after
the termination of pregnancy.
The maternal mortality estimate for Bali of 7.2 per 1000
live births produced by the RAMOS study is much highei' than
the results of the Household Health Survey (1986) for Bali
which counted only 2.3 maternal deaths per 1000 live births.
Part of the difference between these two rates lies in the
fact that the RAMOS questionnaire included all deaths to women
of childbearing age (15-49) and then asked specifically if the
woman had been pregnant or recently delivered at the time of
her death. The Household Health Survey also recorded all
deaths, but did not have a special set of questions on the
pregnancy status of women who had died, and thus obviously
missed some cases which might have been picked up as maternal
deaths under the RAMOS protocol.
3
The prospective study in East Nusa Tenggara (NTT) and
retrospective study in Irian Java recorded deaths by cause,
and births. The Maternal mortality ratios are produced by
relating pregnancy related female deaths to total live births.
In NTT the ratio was 11 deaths to 966 live births, while in
Irian Jaya 4 deaths were recorded compared to 607 live births.
Of course in such studies both deaths and births are subject
to errors of recording and problems of small sample size.
SUMMARY AND RECOMMENDATIONS
Comparisons of maternal mortality estimates in Indonesia
are inhibited by differences of definitions, data collection
methods, sample sizes, and analytical techniques. The need to
carry out research on very large samples, using very careful
procedures for recording mortality and causes makes this one
of the most challenging areas of health and demographic
research, and implies the need for large research budgets and
highly skilled researchers — resources which are
understandably limited in Indonesia.
Data obtained from hospital and maternity clinic records
do not reflect community conditions since most women go to
these facilities only if they are well-to-do (thus biasing MMR
estimates downward) or in need of emergency services (thus
biasing results upwards). In addition deaths of women in the
early stages of pregnancy or in the post-partum period would
not be recorded as maternal deaths in those facilities if they
were not admitted as patients prior to their deaths.
The development of a network of local clinics (PUSKESMAS)
and integrated health service points (POSYANDU) offers the
opportunity to substantially improve the monitoring of
maternal deaths on a nation-wide basis.
In the immediate
future, reports from these units can be used to identify
maternal deaths, which will assist the Department of Health in
developing appropriate intervention programmes.
More
important, the establishment of a large network of health
professionals,
village volunteers, and traditional birth
attendants who have the trust of the community in matters
related to pregnancy and childbirth will facilitate efforts to
improve vital registration records in general. This, in the
long run, is the key to gaining a better understanding of
levels and trends of maternal mortality in Indonesia.
REFERENCES
Alisyahbana, Anna dan kawan-kawan: ’’The Pregnancy Outcome in
Ujung Berung, West Java.
Perinatal Mortality and
Morbidity Survey and Low Birth Weight”, Fakultas
Kedokteran, Universitas Padjadjaran. Final Report V,
p.69, 1983.
4
Budiarso, L. Ratna dan kawan-kawan: "Survai Kesehatan Rumah
Tangga, 1986." Badan Penelitian dan Pengembangan
Kesehatan, Departemen Kesehatan R.I.
Budiarso, L. Ratna, J. Putrali dan J. Prihartono:
"Penelitian
Pengetahuan dan Perilaku dalam Keluarga Berencana serfa
Kesehatan Ibu dan Anak, 1982-1983". Badan Penelitian dan
Pengembangan Kesehatan, Departemen Kesehatan R.I. (n.d.)
"Laporan dan
Budiarso, L. Ratna, J. Putrali dan Muchtaruddin:
Statistik. Survai Kesehatan Rumah Tangga, 1980." Badan
Penelitian dan Pengembangan Kesehatan, Departemen
Kesehtatan R.I.
Budiarso, L. Ratna:
"Pelayanan Kesehatan iba hamil dan
bersalin, Survai Kesehatan Rumah Tangga 1986". Majalah
Kedokteran Indonesia, 38(2): 85-89, 1988.
Chi I-Cheng, Tina Agoestina and Joe Harbin: "Maternal
mortality at twelve teaching hospitals in Indonesia:
kepidemi o1ogic analysis". Int J Gyn & Qbstet, 1981,
19:259-266.
An
Djaja, Sarimawar, L. Ratna Budiarso, Emiliana Tjitra dan S.M.
Pareira: "Gambaran Fertilitas". Lokakarya Sebab
Kematian Bayi dan Penanggulangannya, Kupang, 10-11 Maret
1988.
Fortney, Judith A. et al.: "Maternal Mortality in Indonesia
and Egypt." Interregional meeting on Prevention of
Maternal Mortality, Geneva, 11-15 November 1985.
"Pelaksanaan Survai Kesehatan Rumah Tangga.
Laporan, I.
Propinsi Irian Java 1986". Kantor Wilayah Departemen
Kesehatan, Propinsi Irian Jaya, 1987.
The Safe Motherhood Initiative: A call to
Mahler, Halfdan:
action. The Lancet, March 21, 1987.
Nenobias, T. dan Emiliana Tjitra: "Pola Penyebab Kematian
Umum". Lokakarya Sebab Kematian Bayi dan Penanggulangan
nya, Kupang, 10-11 Maret 1988.
World Health Organization:
International Classi fication of
Diseases,1975 Revision, Vol. 1
5
Research Note Number
30CS
CHILD
SURVIVAL
Date
19 April 1990
Division of Demography and Sociology
Research School of Social Sciences
The Australian National University
Canberra ACT, Australia
A Project Sponsored by The Ford Foundation
IMPLICATIONS OF MORTALITY DECLINE TOWARDS DEVELOPMENT
G.L. Dasvarma
The Population Council,
Jakarta
Child Survival Research Notes are brief discussions of
issues of current relevance to researchers and policy
makers concerned with problems of high infant and child
mortality in the world. The International Population
Dynamics Program, Department of Demography, The Australian
Matronal University, distributes these notes with their
r^^lar Bibliographic Circular. Production of the
V%Child Survival Research Notes is made possible through
"a grhnt from the Ford Foundation. Responsibility for the
)
content of Child Survival Research Notes rests with
the author(s) alone, and not the above-listed organ
isations .
Note:
^ND
1990 - Celebrating Sixty Years of Higher Education in the ACT
ANU Open Day - Sunday 16 September 1990
1.
INTRODUCTION
1.1
Mortality and development
The ultimate goal of development programmes is welfare of the people.
These programmes aim at not only providing the three basic human needs,
namely food, clothing and shelter, but also improving a whole range of
aspects linked to good living such as health, nutrition, education,
education, housing, communications, economic status and so on.
As a part of providing better health and longevity, reductions in
mortality and morbidity are in themselves goals of a development programme.
However, these may also be regarded as means to an end with respect to
other aspects of development, such as better health and longevity helping
to increase productivity, perceived low infant and child mortality helping
to limit fertility, etc.
Reduction of mortality should not be seen merely as a result of
preventing and curing diseases by medical intervention, because the
achievement of a sustainable mortality decline requires concerted efforts
on a wide range of developmental aspects.
As such, the implications of
mortality decline on development should be seen both from the point of view
of costs in achieving a given level of mortality and from the point of view
of the impacts which the given decline in mortality would have on other
aspects of development.
Further, as Hull and Jones (1986) note, it is
difficult to isolate the impact of just mortality or fertility decline
because both are results of social change, improvements in material welfare
and the control of births and deaths made possible by medical and
scientific innovations.
However, since change in mortality is an integral
part of the dynamics of population, it is still possible to consider the
contribution of declining mortality to changes in social and economic
conditions of the people.
1.2
Theory of demographic transition and mortality
It is now well known that the developing countries have experienced a
faster decline in mortality than have the developed countries, in which the
reduction of mortality had to follow the gradual improvements in socio
economic conditions, and innovations in medical and public health
technology.
For the latter group of countries, even though the onset of
mortality decline can be taken to be a precursor to all subsequent
demographic change (Caldwell 1986), there was sufficient scope for
fertility to adjust to the prevailing mortality level, so that there was no
"population explosion".
The developed countries also had another recourse
to alleviate the pressures of a fast expanding population, namely large
scale migration to other sparsely population countries.
In fact, as
Caldwell (1986) reports from findings by Friedlander (1969) , the initial
response to mortality decline in Britain was migration to the United States
and other British colonies around the world and reduction of fertility as a
means of controlling population growth in the face of mortality decline was
postponed.
This option is not available to the developing countries in
the present day and age.
However, within countries, internal migration
can be used as an outlet of population pressure in parts of the country,
such as is being done in Indonesia under the transmigration programme.
The developing countries, on the other hand, having obtained such
improved technology all at one time (though much later), achieved large
scale mortality declines within a short period of time.
The demographic
1
transition began in the developing countries when many of the developed
countries had already completed the main four phases of transition.
In the beginning, without adequate changes in the socio-economic,
behavioural and attitudinal conditions the relatively fast decline in
mortality in the developing countries could not be accompanied by a
corresponding decline in the other component of population growth, namely
fertility (or migration, where applicable) resulting in a "population
explosion" and all its consequences adversely affecting the endeavours
towards economic development and provision of better living conditions for
the people.
In other words, the society was not prepared to absorb the
consequences because the whole spectrum of socio-economic infrastructure
was not conducive to diluting the impact by making efforts to reduce
fertility in a commensurate manner.
This is perhaps the foremost and most
generalized implication one can derive about mortality decline vis-a-vis
development, particularly in the context of developing countries.
However, the present situation tends to be more encouraging as many of the
developing countries including Indonesia have achieved substantial
fertility declines and are continuing on a path towards post-transition
fertility.
The theory of demographic transition focused primarily on the
explanation of fertility decline, leaving a much less attention to the role
of mortality (Caldwell 1982).
Yet a discussion of the role attributable
to mortality is important for at least two reasons - (i) the onset of
mortality decline is often the major determinant of all subsequent
demographic change; and (ii) because so much is the primacy of mortality
decline taken for granted in a fertility-mortality demographic transition
sequence that the assumptions behind the mortality transition theory are
given much less scrutiny than those behind the fertility transition theory
(Caldwell, 1986).
Based on European experience, the causes of mortality
decline can be considered to revolve around the following factors:
economic development and rising living standards; advances in medical
technology and science; improved sanitation and hygiene and increases in
food supply and improved nutrition (Ruzicka and Kane 1989).
However, the
role of each of these factors in terms of timing, mix and balance is still
a matter of debate, particularly in the context of developing countries
(Preston 1978; Ruzicka 1986).
Although, theoretically, population growth can be controlled through
any one of its three components, namely mortality, fertility and migration,
the mortality component is unique in the sense that the reduction of
mortality is the universal aim of governments including those which pursue
policies of slowing down the rate of population growth (Eldridge 1968) .
Unlike changes in fertility, a change in mortality affects all ages
of a population, although the greatest benefits of declining mortality are
derived by infants and children (that is persons aged below 5) followed by
persons aged, say, 60 or 65 and over.
Thus, a declining mortality would
In situations
imply an increasing proportion of under fives and the aged,
of unchanging fertility, a declining mortality would, in the short run
imply a decrease in the crude birth rate of a population because the
denominator becomes larger.
Fertility decline, on the other hand, affects only the base of the
When mortality decline is accompanied by
age pyramid, making it shorter.
a decline in fertility, the population ages both at the base (shrinking of
the base of the pyramid) and at the apex (widening of the apex).
2
2.
CONSEQUENCES OF MORTALITY DECLINE
The consequences of mortality decline can be seen through its various
impacts.
For ease of understanding let us enumerate the different
impacts, such as demographic impacts, impacts on social and cultural
aspects, and impacts on health services delivery.
2.1
Demographic Impacts
2.1.1 Effect on rate of population growth and age distribution
A decline in mortality from an initially high level results in a
large reduction in the crude death rate.
With continuing high fertility
(as was the case in most developing countries soon after the Second World
War), this gave rise to an increase in the rate of population growth.
The
positive or negative consequences of such high rates of population growth
towards development have been subjects of considerable debate, although
there has been an almost general agreement about the negative consequences .
Countries which have perceived the consequences as negative have pursued
antinatalist population policies.
However, at moderate mortality levels,
such as what many developing countries have achieved today, a further rapid
decline in mortality would not produce a very large reduction in crude
death rate, and consequently the impact on the rate of population growth
would not be very big (Population Division 1983) .
Further, as mentioned earlier, a reduction in mortality initially is
Specifically,
concentrated at the earliest ages of infancy and childhood,
with respect to infant mortality decline and its effect on the rate of
population growth, a simple exercise highlighting the relation between the
rates of decline in infant mortality rate (IMR) and crude death rate in
Indonesia shows that, up to a life expectancy at birth of around 65 years,
the rate of decline in IMR is much faster than that in CDR (Dasvarma 1986).
Under constant fertility and constant or no migration, the effect of
mortality decline on the age distribution of a population depends on the
age pattern of mortality change.
A decline from high to moderate levels
produces a reduction in the proportion of population 15 to 65 years or the
segment responsible for economic production, accompanied by an increase
mostly in the 0 to 15 year age segment.
In other words, the (young)
dependency burden would increase.
A further decline from moderate to low
mortality would continue to deflate the proportion aged 15 to 65 years, but
the other age segment to gain in proportion would be mostly that over 65
years, inflating the (old) dependency ratio.
Whereas the developing
countries are concerned more with providing for the 0 to 15 year age
segment in terms of schooling, child care, etc., the developed countries
are concerned to a greater extent with providing for the aged - finding
them support, pension, medical care, etc.
In developing countries,
declines in mortality from moderate to low levels (i.e. from a life
expectancy at birth of around 55 years to higher levels) would lead to
smaller increases in the proportion aged 0 to 15 years.
With a
concomitant decline in fertility, as has been occurring in many developing
countries including Indonesia (due to improvements in socio-economic
factors conducive to both fertility and mortality reduction and due to
explicit fertility control policies followed by the government), the young
age segment would not grow much in proportion, and there will be a larger
growth in the proportion of the old population (Population Division 1983).
3
Such ageing of the population, both at the base and the apex, has
implications for the care of the children and the elderly.
There may be
fewer children due to declining fertility, but new standards of consumption
and schooling costs would force parents to spend more.
There would be
more survivors to old age, who would have to be cared for at costlier
levels of comfort by their children, who would have less number of siblings
(due to reduced fertility) to share the costs of parent care (Hull and
Jones 1986).
2.1.2 Changes in the age composition in Indonesia
The Indonesian population has been undergoing changes in its age
composition as a result of a decline in both mortality and fertility.
Recently Nam et al. (1989) have analysed the past and projected future
* .
To the
changes (to the year 2005) in the age composition of~ Indonesia,
be attributed
to mortality decline (in
extent these projected changes can 1_
-conjunction with a decline in fertility), some of the specific policy
implications suggested in this analysis are worth mentioning here:
- Numbers of children under five will not decline for a long
The numbers
time to come, in spite of declining fertility.
of older children and teenagers will continue to rise.
These will have important implications for educational
policy.
- The number of women of reproductive ages will continue to
increase for a few decades, implying that family planning
budgets will have to be sustained, if not increase.
- The rapidly growing numbers in the working ages would present
a big challenge to economic planners.
More jobs will have
to be creased to keep full employment in the society and to
prevent economic stagnation.
Planners will also have to
consider increasing labour force participation of both men
and women and to link trends in education with working force
needs.
- Although significant ageing problems are not present, nor are
they foreseen in the immediate future, there are signs that
there will be rapidly growing numbers of the aged in the
population.
It would be better to start considering well in
advance about alternative strategies to tackle this problem
so as to avoid the ill experiences of the developed
countries.
- Changes in the age structure can also stimulate changes in
the demand for goods and services, pointing to implications
for the business sector.
2.1.3 Changes in cause of death structure and the age pattern
of mortality
Declining mortality affects the structure of causes of death.
With
immunization and control of deaths due to diarrhoea, acute respiratory
infection and other infectious and parasitic diseases, deaths from such
causes become less dominant.
As a result, chronic and degenerative
diseases such as cardio-vascular diseases and neoplasm begin to dominate
4
the cause of death structure.
The former structure is typical of high
mortality populations whereas the latter is typical of low mortality
populations.
Further, since the chief beneficiaries of mortality control
in high mortality populations are infants and children, the impact of
mortality decline from high to moderate to low levels is to increase the
average age at death.
This process of "modernization" of the cause death structure can be
observed for all developing countries undergoing mortality transition for
which reliable data are available, such as China, the Philippines, Sri
Lanka, Singapore, Peninsular Malaysia, Republic of Korea, Hong Kong, etc.
Further, due to a shift in the average age at death, the typical agepattern of mortality becomes modified from a U-shaped curve reflecting very
high mortality in infancy and childhood to a J-shaped curve (Figures 1 and
2) reflecting a predominance of high mortality risk at more advanced ages
(Ruzicka 1985; Hull and Jones 1986).
Table 1 shows the percentage distributions of deaths by cause in some
selected developing and developed countries and some estimates for Java.
Table 1
Percentage Distribution of Deaths by Cause.
Selected Developing and Developed Countries (late 1970s)
and Java (1972)
Developing
countries
(life expectancy
at birth
55 to 65 years)*
Developed countries
(life expectancy at
birth >70 years)*
Infectious and
parasitic diseases
17
1
29
Respiratory diseases
18
8
17(b)
Accidents and violence
7
7
4
Cardiovascular diseases
18
48
8
Neoplasma
8
21
3
All other causes
32
15
39
Causes of death
Notes:
J ava**
(a) includes TAB, diarrhoea and other infectious diseases,
(b) pneumonia.
Sources: *
Population Division (1984).
Utomo and Iskandar (1986).
The implications of all these is that with the transition from high
to low mortality, the provision of health care services should need to be
reoriented to deal more with the "modern" causes of death.
5
Since a vast majority of deaths in Indonesia are preventable through
the application of simple interventions, the real problem is not one of
finding a cure but of creating the proper economic and organisational
conditions to facilitate the application of available technology.
Further, the dynamic relations among levels of economic development,
mortality levels and patterns of causes of death need to be properly
investigated so that appropriate health interventions suited to current
levels of economic development can be formulated (Hull and Jones 1986).
2.1.4 Effects on morbidity
A reduction in mortality is not necessarily followed by a
corresponding decline in morbidity or improvement in the health situation,
In fact, in the extreme case, it may place an immense burden on a poor
economy by keeping people alive under debilitating conditions (Hansluwka
In high to moderate mortality countries, prevention of child
1984).
deaths from diarrhoea, acute respiratory infections, etc., may put back the
child into the same socio-economic environment and at risk of contracting
another disease (Dasvarma 1984; Hull and Jones 1986).
The implication to
be derived from this point is that efforts to reduce mortality should not
be pursued in isolation, rather it should be integrated with other
developmental efforts which should aim at improving nutrition (thereby
helping to improve immunity against diseases) ; increasing education,
particularly health education and health care behaviour, and at providing
appropriate health education and at providing appropriate health care
facilities.
Based on an analysis of infant mortality in the major Indian
States, Jain (1985) suggests that (female) education and availability of
health services are not interchangeable with respect to their effect on
infant mortality.
In fact, the evidence points to their complementary or
even synergistic roles.
To further emphasize the point about an
integrated developmental approach, one can turn to the impressive mortality
decline achieved by some developing populations such as Sri Lanka, Costa
Rica and Kerala.
The successes in these populations are attributed to
simultaneous improvements in the following factors (Caldwell 1986):
- female autonomy;
- health services, creating universal and easy accessibility,
and efficient working of the system;
- female education equalling that of males;
- universal provision of minimum levels of nutrition;
- universal immunization;
- antenatal and postnatal care.
Meegama (1985), analysing causes of fast mortality decline in
"rapidly declining mortality countries" (RDMC), adds another factor to the
above list, namely sanitary reforms resulting from measures to combat
smallpox and cholera.
The sanitary revolution consisted of providing
piped water and construction of sewers and protected wells.
Palloni
(1985) concludes that the causes of slow decline in the "slowly declining
mortality countries" (SDMC) include a lack of efficacy in the distributive
programmes in the fields of education, health and sanitation.
6
There is a great deal to be learned from the experiences of poor
populations such as Sri Lanka, Kerala and Costa Rica in managing to bring
down their mortality rates when their richer counterparts in other
developing countries still have much higher mortality.
Caldwell (1986)
has identified the ’’routes” which these poor countries have taken towards
low mortality.
One can argue that the fact that these countries have
remained poor is because the resources which otherwise could have been
invested in other developmental activities were spent on provision of
health services and food subsidies.
However, to counter this one can
assert that these poor countries have achieved low mortality by managing
their social welfare and health programme well in spite of their poverty.
2.1.5 Effect on fertility
Declines in mortality can have two types of effects on fertility.
At high to moderate levels and constant age-specific fertility, a decline
in mortality would lead to higher fertility - by increasing the exposure to
pregnancy of women through longer survival of married couples, by
increasing the fecundity through better health and nutrition and by
reducing the proportions of miscarriages.
The crude birth rate, being
influenced by the age distribution, will initially decrease in a mortality
decline from high to moderate levels, because the changes in age
distribution will result in proportionately fewer women in the reproductive
ages and because the denominator would become larger.
However, under constant fertility, declining mortality can also have
a fertility reducing effect by producing longer survival of newborn
children, leading to longer post-partum amennorhoea of breastfeeding women.
Socio-psychologically, when parents start realising that survival of their
children is ensured, they would not need to have a large number of children
in excess of their reproductive goals (Population Division 1983).
Further, within the context of the demographic transition theory, mortality
decline is also associated with changes in the value which parents place on
their children.
Children begin to be no more regarded as providers of
economic and social support through to their (parent's) old age, but become
costly dependents (Okore 1986).
Medical intervention alone cannot produce a sustainable mortality
decline; it has to be accompanied or followed by socio-economic changes.
Therefore, as mentioned earlier, the changing socio-economic, attitudinal
and behavioural conditions which induce mortality decline apart from
medical and public health intervention are also conducive to fertility
decline.
If, in addition to these, an explicit fertility control policy
is pursued, then the fertility inducing effects of mortality decline can be
minimised.
2.1.6 Economic consequences
The economic consequences of declining mortality are generally
regarded as positive (Population Division 1983).
These positive
consequences can be see in terms of increased productivity because of the
lengthening of expected years of working life, improvements in physical and
mental abilities of workers and reduction in absenteeism due to illness
(assuming that declining mortality is accompanied by improvement in
health).
Chernichovsky (1986) suggests that, at least at the household level
declining mortality creates pressures on individuals and households to
7
reallocate resources over the life cycle.
Improved survivorship tends to
induce saving, education and technological change and can also be
considered to be conducive to increased productivity which would help
spread consumption of goods and leisure over a lengthening life span.
2.1.7 Socio-economic differentials
One consequence of declining mortality is a divergence in the socio
economic differentials in mortality.
This is because the higher socio
economic classes and the better educated can gain better access to modern
medical and public health innovations.
In fact, the emergence of
differentials is often considered to be a sign of decline in mortality or
fertility.
But, in a society aiming for equity and welfare, such
differentials cannot be permitted to continue.
A study of causes and
determinants of differentials can provide answers to questions such as how
certain classes have managed to reduce their mortality while others have
not.
It is now well known that maternal education is the single most
important determinant of child mortality.
But the mechanism by which
maternal education operates to reduce child mortality is the lesson to be
learned by societies striving to improve child survival.
Caldwell argues
that educated mothers are better placed to take decisions about their
children's treatment, they can confidently face the modern health care
providers and they can demand treatment when it is not easily forthcoming.
Needless to say, they may be more knowledgeable about disease prevention,
nutrition and proper health behaviour.
Further, as mentioned before,
female education and provision of health care act complementarily and
perhaps also synergistically in (child) reducing mortality (Jain 1985).
3.
ADAPTATIONS TO MORTALITY DECLINE
Adaptation means successful adjustment.
McNicoll (1986) identified
four characteristics of a mortality regime - level and age pattern;
volatility; socio-economic variation; and morbidity and cause of death.
All except volatility have been discussed before, therefore this
characteristic will be briefly explained here.
According to McNicoll
volatility is the periodic fluctuation of death rates.
These fluctuations
are not just the seasonal variations, but results of crisis mortality such
as famine, wars, epidemics and big political upheavals.
High mortality
regimes are much more volatile than low mortality regimes.
In the
European context, the historical change in volatility is attributed to a
gradual shift from epidemic to endemic diseases.
In the case of developing countries such shifts have perhaps
occurred, but more importantly, the efforts of governments and their
capacity to deal with crisis mortality due to famine must have played
significant roles in minimising volatility (McNicoll).
Adaptation to declining mortality by a society can be considered with
McNicoll has suggested
respect to each of these four characteristics,
three modes of adaptation to declining mortality:
(i)
through existing institutional set-up in the society;
(ii)
by eliciting appropriate changes in demographic and
economic behaviour; and
8
(iii) through changes in the larger cultural systems which
impose and preserve a degree of coherence in the meanings
that attach to death and its consequences.
4.
MORTALITY TRANSITION AND HEALTH TRANSITION
The term mortality transition is used to describe the course of
mortality in a population from high (and sometimes fluctuating) to low (and
stable) levels.
Demographers have concentrated on mortality decline
because it can be measured more easily and there is a finality about death
in the sense that it provides indisputable evidence of the failure of
health providing activities.
A broader term than mortality transition is
epidemiological transition, because it includes changes in levels of
sickness as well as mortality.
However, Caldwell (1989) considers that
neither term is sufficient to explain a transition to complete well being,
because both are purely outcome measures.
In searching for a term which
includes social and behavioural changes paralleling the epidemiological
transition and does much to propel the epidemiological transition, Caldwell
and his colleagues have coined a new term called "health transition", which
would include both epidemiological and related social changes.
The new line of study proposed by Caldwell and his colleagues would
aim at examining the cultural, social and behavioural determinants of
health in an attempt to apply the experience of one society's success in
attaining better health to another society.
5.
REVIEW OF RECENT TRENDS AND CURRENT SITUATION
Any discussion of mortality in Indonesia invariably dealt with infant
and child mortality and the life expectancy at birth as implied by the
infant and child mortality.
The estimation of life expectancy is done
from model life tables from the information on infant mortality, although
some authors have shown this not to be strictly valid in view of the
inapplicability of the basic assumptions of the estimation techniques
(Chandrasekaran 1987).
National and provincial level data of somewhat acceptable quality are
available since the 1971 census.
The most recent demographic enquiry was
the 1985 SUPAS.
The recent studies on levels and trends of mortality
include those undertaken by Utomo and Iskandar (1986), Soemantri (1983),
Dasvarma (1986) and Adioetomo and Dasvarma (1987).
5.1
Infant mortality
The infant mortality rate for Indonesia has declined from a high 140
per 1000 live births in the late sixties to 107 in the late seventies
(Soemantri 1983), implying an annual reduction rate of 3.2 per cent.
However, an analysis of the 1976 intercensal survey (SUPAS 1976) produced
an infant mortality rate of 114 (Hull and Sunaryo 1978), which suggested a
slowing down in the pace of decline of infant mortality in the decade of
1970 (Figure 3).
The rate was 4.3% between 1967-68 and 1972-73 slowing
down to 2.8% between 1972-73 and 1977-78.
This appeared to be consistent
with the findings about several other countries by Gwatkin (1980) that the
pace of mortality decline had slowed down.
But the 1985 SUPAS data, with
an infant mortality rate of 70 per 1000 (reference period mid-1984) shows
that the rate of decline between mid-1977 and mid-1984 was more than 8 per
9
cent per annum (Soemantri 1987), that is, if sustained to the end of the
century, more than sufficient to achieve the target set for the year 2000.
The results from SUPAS 1985 imply that the target of an IMR of 70 per 1000
set for the end of REPELITA IV (i.e. by 31 March 1989) has been achieved
almost at the beginning of the REPELITA.
While this may not be totally
impossible, questions still remain as to what might have contributed to
such a fast decline in infant mortality in so short a period.
Much less is known about the causes of death in Indonesia, except for
the two national level surveys conducted by the Ministry of Health in 1980
and in 1985 and a prospective study carried out in West Java in 1982/83.
Even these surveys provide information on causes of death mostly among
infants (under 1 year of age) and children (aged between 1 and 5 years).
According to these studies, between 25 and 30 per cent of deaths below five
years of age occur in the neonatal period (first month after birth) and
another 30 per cent in the period 1 month to five years of age.
About 47
per cent of all deaths under 5 was due to infectious and parasitic diseases
(measles, diphtheria, pertusis, tetanus, typhoid, cholera, hepatitis).
A
large proportion (41 per cent) of neonatal deaths (first month after birth)
was caused by tetanus alone.
Most of the deaths cited above are
preventable.
Thus, if a good and effective programme of immunization
against tetanus, diphtheria, whooping cough, polio and measles, and proper
delivery of children is undertaken, infant mortality would be reduced
considerably.
1
Vigorous activities in the areas of (i) maternal and child health
(MCH), (ii) nutrition, (iii) immunization, (iv) control of diarrhoeal
disease and (v) family planning (the five key programmes in health) started
only after 1980 and implemented as an integrated programme at the village
level (POSYANDU) from 1985 (Utomo and Iskandar 1986, p.87).
Given the
relative recency of these programmes, it does not seem highly likely that
there will be such a large scale reduction in infant mortality between 1980
and 1985.
However, mortality, like fertility is very much influenced by
socio-economic and behavioural conditions.
These conditions may very well
have assumed a stage conducive to reduced mortality, and only seemingly
small health interventions could have been necessary to achieve a large
It is imperative that a study be done which
scale mortality reduction,
could throw light on the factors associated with mortality decline in
Indonesia, especially in the last decade or so.
5.2
Prospects for further decline
At high levels of infant mortality (30 per 1000 live births and
above), there is a large percentage of deaths of infectious origin which is
amenable to medical intervention.
In this situation, effective
implementation of medical and health programmes can reduce infant mortality
relatively easily.
This level of infant mortality has been compared by
Bourgeois-Pichat (1978) with the "Soft Rock" part of soil which can be
eroded away easily to reveal the hard rock.
At infant mortality below 30
per 1000 live births there are few deaths of infectious origin and a large
percentage of deaths is due to congenital defects and cardio-vascular
diseases; Bourgeois -Pichat has compared this with the "Hard Rock".
However, D'Souza (1984) has noted that in the developing countries, where
the infant mortality rate is 100 infant deaths or less per 1,000 live
births, any further reduction in the rate cannot be achieved through
medical intervention alone - simultaneous improvements in social structure
and environmental health are also needed.
The importance of the latter
increases as infant mortality decreases further.
At levels of infant
10
mortality below 30 per 1,000 live births further declines cannot be
effected without significantly improving birth processes and controlling
lifestyle-related deaths.
Therefore D'Souza has proposed a
reclassification of Bourgeois-Pichat's dichotomy of "soft rock" and "hard
rock" into three categories, namely "soft rock" (infant mortality 100 or
more per 1,000 live births), "intermediate rock" (infant mortality between
100 and 30 per 1,000 live births) and "hard rock" (infant mortality below
30 per 1,000 live births).
Currently Indonesia can be regarded as
belonging to the "intermediate rock" category, therefore action is required
both on the medical/health front and the socio-economic front in order to
achieve further declines in infant and child mortality.
5.3
Adult Mortality
For obvious reasons, such as lack of vital statistics, and the
inapplicability of indirect methods of estimating adult mortality, the
information on age-specific death rates (comprising adult as well as infant
and childhood ages) is practically non-existent in Indonesia.
Therefore,
the sole mortality measure that includes adult mortality is the life
expectancy at birth, estimated from model life tables on the basis of
infant mortality rates.
The only study to date providing age specific death rates is that
based on the Sample Vital registration Project (SVRP) carried out by the
Central Bureau of Statistics between 1974 and 1977 in nine Kecamatans and
Jakarta (Gardiner 1978).
This study showed that mortality in late
childhood and early adulthood (after J0-40) was much lower than that in
Further, female mortality after age 60 was
infancy and late adulthood,
These findings are contrary to the patterns
higher than that of males.
adopted in the model life tables, which are the bases for estimating
expectations of life at birth from estimates of infant mortality.
It
points to the need for collection of appropriate data for better estimation
of mortality in late childhood, adult ages and old ages, separately for
males and for females.
Ideally, these should be based on direct
information such as vital statistics; but even data collected separately
for males and females through censuses or surveys could be very useful in
filling the gaps.
5.4
Cause of death structure
With the eventual modernization of the cause of death structure in
Indonesia different strategies will have to be adopted to deal with the socalled diseases or causes of death of "civilization" such as cancer and
heart diseases.
Furthermore, newly imported diseases such as AIDS
(acquired immuno deficiency syndrome), though not causing immediate
concern, are worth being vigilant about.
With prosperity, more and more persons are owning and driving motor
vehicles.
Although statistics are lacking, it is not difficult to imagine
that deaths due to road traffic accidents are assuming significant
proportions.
Enforcing speed limits on highways and using seatbelts could
reduce such fatalities.
Wearing safety helmets for motorcycle riders has
been made compulsory, but not all wear them properly.
5.5
Differential mortality
Considerable differentials by rural-urban residence and by province
have been observed in Indonesia (Soemantri 1983, 1987).
The provinces
11
with the lowest IMR around 1971 have achieved the fastest declines by 1980
or 1985, so that the magnitude of the differentials have remained more or
less similar during this 10/15 year period (Figure 4).
One means of
reducing infant mortality at the national level is to reduce the inter
provincial or other differentials.
The differentials in infant and child mortality in Indonesia exhibit
similar patterns as observed elsewhere.
It shows a U-shaped pattern with
respect to mother's age, i.e. babies born to very young and very old
mothers have a higher risk of dying than those born to adult women
(Rutstein 1983, p.74).
Order of birth influences the child's chances of
survival, particularly if considered in conjunction with mother's age;
children born to women aged 25-29 with low parity (one or two) have been
found to have the highest survival chances (Kadarusman 1982).
Children
born after shorter birth intervals have higher mortality, while male
children have a considerably higher risk of death than female children
(Rutstein 1983, pp.74-75).
With respect to socio-economic factors, the well-known relation of
maternal education to infant and child mortality holds true also for
Indonesia as revealed by data from the 1971 census (Hull and Sunaryo 1978)
and the 1980 census (Adioetomo 1983).
Economic status of a household as
measured by floor area also shows the expected inverse relation to infant
and child mortality (Adioetomo 1983).
6.
SUMMARY AND CONCLUSIONS
The general implications of mortality decline towards development
will be discussed in this section with special reference made to
implications for health planning and policy.
6.1
Implications for development in general
A reduction in the death rate of a population would result in a
change in one or more of the following demographic parameters: birth rate,
net migration rate or rate of growth of total and per capita products.
The demographic response to a decline in mortality may be lower fertility
or higher net out-migration; the economic response may be increased output
to prevent a lowering of per capita production (McNicoll 1986).
Planners
and policy makers who deal with mortality reduction targets in terms of
crude death rates or life expectancy at birth may be advised to note that
as mortality declines from moderate to low levels, gains in life expectancy
will become smaller (Schoen 1986).
The crude death rate, being a function
of the age structure of a population will likewise show smaller increments
and may even actually increase.
That is why it may be incorrect to set
mortality targets simply in terms of the crude death rate (Dasvarma 1988).
As mentioned before, declining mortality, coupled with declining
fertility would lead to ageing of the population, both at the base and the
apex of the age-sex pyramid, implying that in course of time there will be
similar numbers of people in successive age groups.
These changes would
put pressure on society to alter age-old institutions governing production,
exchange, leadership and socialization which were developed to cope with
high mortality-high fertility situations.
Systems of care for the
children and the elderly would be under strain.
Parents would be under
pressure to spend more (i) on their children (though less in number per
family than in previous generations) to meet new standards of consumption
12
including schooling, and (ii) on their own parents, who would live longer
and become accustomed to higher standards of comfort due to concomitant
improvements in the economy (Hull and Jones 1986).
An analysis of projected changes in the age composition of the
Indonesian population points to specific implications for planning for
education, employment, links between education and employment, family
planning and for the business sector to make provisions for changing demand
for goods and services.
6.2
Implications for health planning
Paradoxical though it may sound, declining mortality can create
greater demands on health services.
First, although fatality may decrease
through prevention (e.g. immunization) or cure (e.g. oral rehydration
therapy for diarrhoea), morbidity may not because the child or adult may
continue to live in the same environment chacterised by poor nutrition,
poor sanitation, low levels of education and inappropriate health
behaviour.
Secondly, mortality decline from moderate to low levels,
accompanied by increase in education and awareness may change the
individual's perceptions of illness.
With an increase in perceived
illness, individuals would tend to seek medical attention more often,
thereby inflating the cost of health care.
This is perhaps one reason why
in the more developed countries low mortality levels do not necessarily
mean low morbidity and hence the non-fulfilment of the "health transition".
The other paradoxical implication of mortality decline towards health
care stems from changes in the structure of causes of death or of illness.
A shift from predominantly infectious and parasitic disease structure to
one dominated by chronic and degenerative diseases will necessitate a
change in the orientation and emphasis of health care provision.
Even under current levels of mortality and health, it has often been
demonstrated that health care modelled on the Western system fails to
achieve the desired results, because communities with their own traditional
systems are at conflict with the government provided Western based systems.
That is why it may be wise to evolve a system whereby the traditional and
the modern methods are integrated.
The importance of a concerted effort on several fronts to improve
longevity and health can be demonstrated by an example from Mosley (1983):
Consider two major killer diseases of childhood, measles and diarrhoea.
Before appropriate technologies were available (in the West before the
start of mortality transition) mortality control was achieved primarily by
social changes relating to improved nutrition and family hygiene.
The
effect was to reduce the incidence of diarrhoeal diseases (by reducing the
frequency of diarrhoea) and to reduce the severity of measles (by improving
host immunity through improved nutrition).
The present technological
approach to primary health care concentrates on measles immunization to
reduce the incidence of the disease, not necessarily accompanied by an
increase in host resistance (if nutrition has not improved).
With respect
to diarrhoea the present approach of oral rehydration therapy is expected
to reduce the severity of the disease without reducing its incidence.
13
/
i
REFERENCES
Adioetomo, Sri Moertiningsih (1983): Diferensial mortalitas bayi dan anakanak di Jakarta dan Indonesia - analisa data sensus penduduk
Indonesia 1980 (Differential infant and child mortality in Jakarta
and Indonesia - analysis of the 1980 Indonesian population census
data). Demographic Institute, Faculty of Economics, University of
Indonesia, Jakarta.
and G.L. Dasvarma (1987): Levels and trends of
child mortality in Indonesia based on the 1980 census. Third National
Statistical Congress, Jakarta, 16-28 June 1987.
Bourgeois-Pichat, J. (1978): Future outlook of mortality decline in the
world. Population Bulletin of the United Nations, No.11, New York.
Caldwell, John C.
London.
(1982): Theory of fertility decline. Academic Press,
(1986): Routes to low mortality in poor countries,
Population and Development Review, 12(2):171-220.
(1989): Introductory thoughts on health transition. Paper
presented at the Health Transition Workshop, The Australian National
University, Canberra, 15-19 May 1989.
Chandrasekaran, C. (1987): Levels of fertility and mortality in Indonesia
in 1985 and prospects for their further decline. Report prepared for
the Ministry of Population and Environment, Jakarta.
Chernichovsky , Dov (1986): Interaction between mortality levels and the
allocation of time for leisure, training, consumption and saving over
the live cycle. In United Nations (1986), pp.126-131.
Dasvarma, G.L. (1984): Infant and child deaths in Indonesia - how many can
be prevented? Majalah Demografi Indonesia, 1984.
(1986): Infant mortality and crude death rate. Memorandum
submitted to the Assistant Minister for Population, State Ministry of
Population and Environment, Jakarta.
(1988): Mortality targets for REPELITA V. Communication to
KLH, Jakarta 1988.
D'Souza, Stan (1984): Measures of preventable deaths in developing
countries - some methodological issues and approaches. IUSSP Seminar
on Social Correlates of Mortality, Tokyo, 24-27 November 1984.
Eldridge, Hope T. (1968): Population Policies. Cited in United Nations
(1973) Determinants and Consequences of Population Trends, New York,
ST/SOA/Ser.A/50.
Friedlander, Dov (1969): Demographic responses and population change,
Demography, 17(4):359 - 381.
Gardiner, Peter (1978): Age patterns of mortality in Indonesia. Working
Paper No.12. Indonesian Population Dynamics Project, Population
Studies Center, University of Gadjah Mada, Yogyakarta.
14
Goldman, Noreen (1986): Effects of mortality level on kinship.
Nations (1986), pp.79-81.
In United
Gwatkin, Davidson R. (1980): Indications of change in developing country
mortality trends - the end of an era? Population and Development
Review, 6(4).
Hansluwka, H.
(1984): Mortality in South East Asia - a Review.
Hull, Terence H. and Sunaryo (1978): Levels and trends of infant and child
mortality in Indonesia. Population Studies Centre, Gadjah Mada
University, Yogyakarta.
International
and Gavin W. Jones (1986): Introduction
In
United
Nations
(1986), pp.1mortality trends and differentials.
9.
Jain, Anrudh K. (1985): Relative role of female education and medical
services for decreasing infant mortality in rural India. In Halstead,
Scott B., Julia Walsh and Kenneth S. Warren (eds.): Good Health at
Low Cost. Proceedings of a Conference held at the Bellagio Conference
Center, Italy, April 29-May 3 1985, The Rockefeller Foundation, New
York, 1985.
McNicoll, Geoffrey (1986): Adaptation of social systems to changing
mortality regimes. In United Nations (1986), pp.13-19.
Meegama, Anandar (1985): The mortality decline in the fast declining
developing countries , International Population Conference, Florence
1985, Proceedings, 2:317-328, IUSSP.
Mosley, W. Henry (1983): Primary health care - rhetoric and reality, Populi
- Journal of the United Nations Fund for Population Activities,
10(3):41-53.
Nam, Charles B., Sri Pamoedjo Rahardjo and Gouranga Dasvarma (1989): The
changing age distribution in Indonesia and some consequences. Paper
presented at the Special Meeting on Indonesian Population, Baltimore,
Maryland, March 17-28, 1989.
Okore, Augustine 0. (1986): Effect of changing child mortality on value of
children to parents. In United Nations (1986).
Palloni, Alberto (1985): An epidemic-demographic analysis of factors in the
mortality decline of slow decline developing countries, International
Population Conference Proceedings, 2:329-352, IUSSP, Florence, 1985.
Population Division (1983): Update - Mortality and health policy, Populi Journal of the United Nations Fund for Population Activities, 1983
pp.6-30.
Preston, S.H. (1978): Mortality, morbidity and development. Population
Bulletin of the United Nations Economic Commission for Western Asia,
No.15.
Rutstein, S.O. (1983): Levels, trends and demographic differentials, WFS
Comparative Studies, No.24, World Fertility Survey, London.
15
Ruzicka, Lado T. (1985): Causes of mortality change - observations based on
the experience of selected countries in the ESCAP region. In
Mortality and Health Issues: Review of current situation and study
guidelines, Asian Population Studies Series No.63, ESCAP, Bangkok,
1985.
(1986): The elusive path of mortality transition. In
Hansluwka, H. et al. (eds.), New Developments in the Analysis of
Mortality and Causes of Death. World Health Organisation, Geneva, and
Mahidol University, Bangkok
and Penny Kane (1989): Health transition - the course of
morbidity and mortality. Paper presented at the Health Transition
Workshop, The Australian National University, Canberra, 15-19 May
1989.
Schoen, Robert (1986): The direct and indirect effects of mortality decline
on demographic variables. In United Nations (1986), pp.20-30.
Soemantri, S. (1983): Pola perkembangan dan perbandingan antar daerah angka
kematian bayi (Trends and differentials in infant mortality among the
provinces of Indonesia). Paper presented at a seminar on infant
mortality in Indonesia, Central Bureau of Statistics, Jakarta, 1-3
February 1983.
(1987): Angka kematian bayi di Indonesia - tingkat,
kecenderungan dan prospeknya (Infant mortality in Indonesia - levels,
trends and prospects). Paper presented at the third National
Statistical Congress, Jakarta, 26-28 June 1987.
United Nations (1986): Consequences of mortality trends and differentials ,
New York, ST/ESA/SER.A/95.
Utomo, Budi and Meiwita B. Iskandar (1986): Mortality transition in
Indonesia 1950-1980 , Asian Population Studies Series No.74, ESCAP,
Bangkok.
16
Fijun 11
,, ,
, ,,,
, .
Patterns of age-specific fenale death rates (Mx) for high Mortality countries,
i,N, "South Asia" Model life tables,
(Deaths per 108,008 population)
It
8,1-
Death Rate Ol
0,881-
8,8081 -- 1—M- 1-- 1- H—1-- 1-- 1-- 1-- 1-- 1-- 1-- 1- 1-- *-- •
8 5 18 15 28 25 38 35 40 45 50 55 68 65 70 75 88 85
ftye (lean's)
-*-e(0): 48 years
-+- e(0): 58 years
-*-e(0)60 years
Patterns of aye-specific Hale and female death rates (Mx) for low mortality countries.
Coale and Demy "Nest" model life tables,
(Deaths per IMl population)
1000T
180Males (e(0)
years)
-+■ Females (e(0)
75 years)
Death Rate 18-
8.1
4 -- 1-- 1-- 1—4-- 1-- 1-- 1-- 1-- 1-- 1-- 1-- 1-- *-- H—I-- 1
5 18 15 28 25 38 35 48 45 58 55 58 65 78 75 88 85
ftye (Years)
•r«..
Figure 3:
Trends in infant Mortality rate (IMR) in Indonesia
150t
125-
108-
Mortality 75Rate
504251975
Year
•
.•
'''
v’ ’
.■
;
Si
Ss
Figure 4:
Trends in nfant Mortality rate (INB) provinces Nith the highest and lowest rates
and the national average
225 t
NIB
200-
175150-
Indonesia
Mortality
Rate 100-
Yogyakarta
75-
50-
25965
1970
1975
Year
1980
1985
Research Note on
CHILD
SURVIVAL
Number
Date
23 cs
16 March 1989
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
MOBILE PAEDIATRIC CLINIC SERVICES AND CHILD MORBIDITY
IN A RURAL HEALTH PROJECT AREA IN INDIA
H.N. Ranganathan
and
L.D. Puranik
King Edward Memorial Hospital Research Centre,
Pune, India
Note:
(-
ano
X
>
'Nt°^ArION
/y
&
.
Child Survival Research Notes are brief discussions of
issues of current relevance to researchers and policy
makers concerned with problems of high infant and child
mortality in the world. The International Population
Dynamics Program, Department of Demography, The Australian
National University, distributes these notes with their
^regular Bibliographic Circular. Production of the
Iphild Survival Research Notes is made possible through
'a. grant from the Ford Foundation. Responsibility for
the content of Child Survival Research Notes rests with
the author(s) alone, and not the above-listed organ
isations .
INTROpyCTION
The Vadu Rural Health Project, started in the year 1977 by the King
Edward Memorial Hospital (KEM), Pune, which is a non-Governmental voluntary
institution, has been concentrating its efforts in operationalising the
Primary Health Care approach for achieving the goals of ’’Health For All by
the Year 2000”.
This project serves at present a population of about
40,000 spread over 19 villages and is located in Pune District, Maharashtra
State, India.
The location maps of the project are shown on Pages 8 and
9.
It has often been observed that the people living in far-off villages
find it very difficult to avail themselves of services provided at a rural
health centre or hospital, especially child health care, due to
accessibility.
Therefore one of the innovative approaches experimented
with in this project in the Maternal and Child Health Programme is to
provide specialists’ services to the rural people in their own villages
which are away from the village in which the Primary Health Centre is
located, through frequent and regular mobile clinics conducted by a
paediatrician with the assistance of a Medical Social Worker.
This paper
intends to study the utilisation of the mobile paediatric clinic services,
and also to ascertain the leading causes of illness noticed by the
paediatrician amongst the children.
Community-based studies of this type,
which are important from the point of view of accessibility, provision and
utilisation of specialised health services in the rural areas, are scanty.
It is hoped that the findings of the study would be useful in Community
Health Programmes of developing countries for providing such regular mobile
specialists’ services to the villages which are far off and do not have
such facilities.
METHODS AND.. MATERIAL.
The Mobile Clinic functions as follows:
(i)
Prior intimation to the people of the concerned village
about the date, time and place of the clinic through the
Multipurpose Worker or Community Health Guide o± the
village to facilitate the maximum number of children to
avail of the clinic facility;
(ii)
Clinical examination of children by the Paediatrician;
(iii) Treatment and advice;
(iv)
Health education taking advantage of the personal and
close contact between the patient (or parents - mostly
mothers of the children) and the doctor;
(v)
Referral services, wherever necessary.
1
The mobile team* draws up a plan of its visits for the whole year in
advance and maintains a clinic register in which the particulars of every
child attending the clinics are systematically recorded on the following
items:
(i)
Name of the village.
(ii)
Date of clinic.
(iii) Name of the child or parents in the case of newborns and
infants.
(iv)
Age.
(v)
Sex.
(vi)
Illness noticed by the Paediatrician, after ascertaining
the history of illness and examination of the child.
(vii) Treatment and advice given.
(viii)Referrals.
The information furnished on the above items and brief clinical notes
provided the primary data for the purpose of the study.
The study covered a twelve-month period of the calendar year 1986 and
it relates to 18 out of 19 villages of the project area as the remaining
one village, viz. Vadu Budruk, has a rural hospital with a bed strength of
30.
FINDINGS
Clinics held in....villages
The number of clinics conducted by the mobile team in the villages
varied from 4 to 7 with an average of 4.8 per village.
This means,
generally, a clinic was held once in two or three months.
Although the interval between clinics was rather long, the mobile
team has the potential of conducting the clinics at shorter intervals with
the demand for such services increasing.
The villages have been classified into two main groups - interior and
roadside.
The latter are relatively better than the former from the
communication point of view.
It is seen from Table 1 that slightly more
clinics were held in roadside villages compared to interior villages.
* The headquarters of the mobile team is in Pune, which is a large city and
also the headquarters of Pune district.
2
Table 1
Average . . Number of Clinics;. . . conducted per village
Type of
village..
No. of
.villages.
Total clinics
conducted.
Average No. of
.clinics per village
Roadside
12
60
5.0
Interior
6
27
4.5
All villages
18
87
4.8
C.U.nic_.atte^^
The daily attendance of children in the clinics varied from as low as
7 to as high as 51, averaging to 20.8. Repeat
Repeat cases
cases were
were negligible,
negligible. In
the neighbouring Primary Health Centres, the average daily outpatient
attendance (including adults and children) is reported to be about 15.
Thus the attendance of children in the mobile clinics conducted in their
own villages is considered satisfactory, and the utilisation of services is
quite encouraging.
The clinic attendance was better in roadside villages
- 21.4 as compared to 19.3 in interior villages.
The average population
of an interior village is 945 (1981 census) as against 2055 of a roadside
village which suggests that the population of children also in the latter
is more than in the former.
Perhaps due to this population differential,
the clinic attendance was more in roadside villages.
Table 2
Average. Daily Attendance of Children in the Clinics
Average
attendance
___ per.clinic
Type of
village
No. of clinics
conducted
No. of children
attending the
Roadside
60
1286
21.4
Interior
27
522
19.3
All villages
87
1808
20.8
Sick children
The data in Table 3 show that Sick Children formed 60 per cent of the
children attending the clinics.
This percentage is considerably higher in
interior villages compared to roadside villages - 67.0 and 60.6 per cent
respectively.
3
It is normal to expect a high percentage of sick children in a clinic
or hospital because most of the parents bring their children with minor or
major illness to the clinic for relief.
However, it is observed that not
only sick but also healthy children (37.6 per cent) were brought to the
clinics by the parents for general medical check-up and advice.
This
changing attitude of the rural community should be considered as a good
sign of utilisation of the clinic services.
Table 3
Percentage of Sick Children Attending the Clinic by^Sex
Total children
„„aXtendijng. clinics
Sick children
% children
who... were.sick
Male
Female
966
842
608
521
62.9
61.9
Total
1808
1129
62.4
Male
Female
662
624
402
377
60.7
60.4
Total
1286
779
60.6
Male
Female
304
218
206
144
67.8
66.1
Total
522
350
67.0
Sex
All villages
Rpadside.j7il.lages
Interior villages
Mor bidi ty. causes
In order that every village was equally represented and every season
was covered, the records of four clinics only, each corresponding to one
season in respect of every village were studied for ascertaining the causes
of children’s illnesses.
As far as possible, the clinic conducted in the
middle month of the season was selected.
The seasons are:
(a)
December, January and February;
(b)
March, April and May;
(c)
June, July and August;
(d)
September, October and November.
Thus, the primary data for ascertaining the morbidity causes of
The number of children
children consisted of the records of 72 clinics.
4
who attended the clinics was 1014 of whom 638 were found to have at least
one ailment at the time of examination by the Paediatrician.
The specific diseases which accounted for at least one per cent of
sick children are shown in Table 4.
The first five leading diseases diarrhoea, acute respiratory infections, anaemia, otitis media and
The
conjunctivitis - are common to both infants and children aged (1-4).
other leading causes and their ranks are not the same from rank 6 onwards
in the two age groups.
It is also seen that diarrhoea and acute
respiratory infections account for very high morbidity amongst infants as
well as children aged (1-4).
These two diseases are also reported to be
the main causes of childhood mortality in developing countries (WHO/UNICEF,
1986), and therefore their control needs priority attention so that there
will be substantial improvement in child survival.
Table 4
Lead.in£....Causes.of.Illness.Among.Children
........ChildrenAged.Ll-4LAQars
.Infants.(.<Uearl
Rank
Diseases
Rank
Diseases
1
Diarrhoea (34.6)
1
Diarrhoea (33.2)
2
Acute respiratory
infections (31.2)
2
Acute respiratory
infections (11.5)
3
Anaemia (5.7)
3
Anaemia (7.5)
4
Otitis media (4.3)
4
Otitis media (6.1)
5
Conjunctivitis (2.7)
5
Conjunctivitis (5.6)
6
Umbilical sepsis (2.4)
6
Pica with mention of
iron deficiency (4.2)
7
Impetigo (1.4)
7
Pica without mention
of iron deficiency (3.8)
8
Umbilical hernia (1.0)
8
Avitaminosis, protein
calorie deficiency (3.8)
9
Avitaminosis, protein
calorie deficiency (1.0)
9
Impetigo (3.5)
10
Measles (3.1)
11
Worm infestations (2.1)
12
Umbilical hernia (1.4)
Note :
Figures in parentheses are percentages of sick children due to the
stated illness among the total sick children of the respective age
groups.
Multiple classification of sick children has been done, if
more than one cause was mentioned in the clinic registers.
5
Pisea^.s...cpyered.under.the.Universal. ImmunizaUP^^
. LUI.PJ.
Amongst the 1014 children seen in the 72 clinics, there were no cases
due to whooping cough, diphtheria, tetanus and poliomyelitis.
However, 11
children were found to be suffering from measles and 3 children had
tuberculosis.
In the project area, measles was not included in the past
in the immunization programme, and probably for this reason a considerable
number of measles cases had occurred.
It has now been included under the
UIP.
PbY.si.cal.ly disabled and mentally retarded children
Eleven physically handicapped and 9 mentally retarded children were
detected in the clinics, and their rates are 10.8 and 8.9 per 1000 children
attending the clinics.
Out of the 11 physically handicapped children, 6
were deaf and dumb, 3 were blind (including partial blindness) and the
remaining 2 were lame.
Nutri ti onal.st atus
Children were graded for their nutritional status following the Gomez
classification.
It is seen from Table 5 that the percentage of children
with ’’Normal” nutritional status is much higher for males compared to
females.
The observed sex differential might be due to the preferential
treatment given to male children by parents in developing countries,
particularly in the rural area.
Table 5
Nutritional
Nutritional
_M...81 a t u s. —
Children by Age...and. Sex
(%)
Age in Years
0~
... J,-4
0-4___
Males
Normal
Grade I
Grade II
Grade III
80.6 (187)
12.1 (28)
4.3 (10)
(7)
3.0
70.2 (217)
16.5 (51)
11.0 (34)
(7)
2.3
74.7 (404)
14.6 (79)
8.1 (44)
2.6 (14)
Total
100.0 (232)
100.0 (309)
100.0 (541)
Normal
Grade I
Grade II
Grade III
70.8 (170)
15.0 (36)
10.0 (24)
4.2 (10)
56.2 (131)
16.3 (38)
10.9 (44)
8.6 (20)
63.6 (301)
15.6 (74)
14.4 (68)
6.4 (30)
Total
100.0 (240)
100.0 (233)
100.0 (473)
Females
Note:
Figures in parentheses are the number of children
in each nutritional status and age group.
6
Referral_^ryices
The nearest hospital to which referrals can be made is the rural
hospital at Vadu Budruk village.
The next higher level hospital is the
KEM Hospital, Pune, or Sassoon General Hospital, Pune, attached to the
Government Medical College.
These hospitals are approachable from the
villages of the project area by road.
However, in rainy months, the
interior villages face some difficulty as the roads are bad.
Amongst the 1014 children seen in the clinics, 75 were referred to
the Vadu Rural Hospital or KEM Hospital, Pune, for detailed examination and
treatment.
Thus referral services were given to 7.4 per cent of the
children examined.
Cases were referred from all the villages of the
project area and most of them were referred to the Vadu Rural Hospital.
The important causes for which referrals were made were.- severe diarrhoea,
otitis media, acute respiratory infections needing hospitalisation,
premature delivery, tuberculosis, conjunctivitis, congenital abnormalities
and mental growth retardation.
CONCLUSION.
It can be said that the utilisation of the mobile paediatric clinic
services which have been provided to the rural people almost at their
doorsteps, has been encouraging.
Besides, the rural community has the
satisfaction of having the regular services of a specialist easily
The study shows that more attention to the interior
accessible to them,
villages is needed.
The pattern of morbidity causes of children in rural India may not be
different from the one observed in this project area, and by giving
priority attention to these basic problems, viz. diarrhoea, acute
respiratory infections, nutrition and also through intensive health
education, substantial improvement in the health of children can be
expected.
ACKNOWLEDGEMm
The authors are grateful to Dr. (Mrs.) Banoo J. Coyaji, Chairperson,
and Dr. V.N. Rao, Director, KEM Hospital Research Centre, Pune, for
permission to publish the findings of the study.
REFERENCE
WHO/UNICEF, 1986, Basic principles for.^control of acute respiratory
infect ions ...in children in de.yelo.ping countries, 58pp.
7
MflP 1
LOCATICN OF VADU PUPAL HEALTH PROJECT AREA IN MftHARfiSHTRA AND
_
IN INDIA
MAHARASHTRA
VADUPROJGCF
AREA.
INDIA
MAHARASHTRA
VAOU
.
8
MAP 2.
LOCATION OF VILLAGES COVERED UNDER THE VADU RURAL HEALTH PROJECT
xTATEGAoh
0
3k.
•w
PIN1PALE. .
^Cf TAC^TAP
VAJE
WADI
50
Os IKRAPUR^
>3AT4ASVMD|
VADU ,
• APTI
) ©
tulapuk
O
/xM
-
PtWLGAON
VADD
® KH-
O KOR^MON
CH IMA
lonikai
O/^
LEGEMD
P£RAN£
£» Hanor^p
0
o sobcentrje
o
e BHAUAD1
DOHGAR.Gt^ON
■VADU R.URAL -HQ^PnAU
• OTHER. VILLAGES.
e BAkori
X
9
Research Note on
CHILD
SURVIVAL
Number
Date
24CS
20 April 1989
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
FACTORS AFFECTING CHILDHOOD IMMUNIZATION
IN RURAL BANGLADESH
Nltai Chakraborty,
Department of Statistics,
Dhaka University, Bangladesh
and
Kim Streatfield,
Child Survival Project,
Department of Demography,
The Australian National University
Note:
•j
M <
1
'-/VO
C£^Tfe
Child Survival Research Notes are brief discussions of
issues of current relevance to researchers and policy
makers concerned with problems of high infant and child
mortality in the world.
The International Population
Dynamics Program, Department of Demography, The Australian
National University, distributes these notes with their
regular Bibliographic Circular.
Production of the
Child Survival Research Notes is made possible through
a grant from the Ford Foundation.
Responsibility for
the content of Child Survival Research Notes rests with
the author(s) alone, and not the above-listed organ
isations .
In the developing world about one third of child deaths
can be directly attributed to four diseases - neonatal
tetanus, pertussis (whooping cough), measles, and acute lower
respiratory tract infections.
For example, in a surveillance
area in Matlab, Bangladesh, 35 per cent of the child deaths
under age 5 were attributed to these diseases (Chen et al.,
1980:27).
In Bangladesh tetanus alone accounts for 223,000
child deaths and measles accounts for 20,000 to 40,000 deaths
every year (Government of Bangladesh, 1985a:l).
Other
immunizable diseases, whooping cough, polio and tuberculosis,
also contribute significantly to the child deaths.
Prevention of these diseases through immunization would save
millions of lives every year in the developing world.
To reduce the morbidity, disability and mortality
associated with these vaccine-preventable diseases, the
Expanded Program of Immunization (EPI) was established by the
World Health Assembly in 1974 and set the long-term global
objective of reducing morbidity and mortality from six major
diseases - diphtheria, whooping cough, tetanus, measles,
polio, and tuberculosis - by immunizing at least 85 per cent
of children throughout the world by the year 1990.
An
extensive immunization campaign in developed countries and in
some developing countries has brought about a dramatic decline
in death rates due to the immunizable diseases.
However,
recent estimates (excluding China) are that at most 40 per
cent in the developing world are protected against diphtheria,
whooping cough, polio and tetanus and only 20 to 30 per cent
are protected against measles (Population Information Program,
1986:L154).
In Bangladesh EPI was formally launched in 1979
and since then there has been a steady increase in the number
of centres offering immunization services.
It is now
estimated that 22 per cent of the population has access to
immunization services, but the actual practice of immunization
remains very low.
Government immunization service statistics
for 1984 show that the national coverage for BCG was 1.5 per
cent, for DPT (3 doses) 1.4 per cent, for Polio (3 doses) 1.1
per cent and for Measles 0.9 per cent (Government of
Bangladesh, 1985b).1
Socio-Demographic Factors
A number of studies of the utilisation of health care
services have found a relationship between utilisation
behaviour and socio-economic and demographic factors, such as
age, sex, education, family size, occupation, religion and
ethnicity.
A study of immunization of pregnant mothers with tetanus
toxoid vaccines in a rural area of Bangladesh showed that
acceptance of tetanus immunization varied with differences in
the socio-demographic characteristics of the population
(Rahaman et al., 1982:272).
It was observed that as compared
T Based on reports received by the World Health Organization
as of January 1989, coverage estimates for these vaccines
are probably around 50 per cent of the developing world, but
still only 11-23 per cent for Bangladesh (EPI, 1989).
1
with non-acceptors, acceptors were more often Hindus, had a
higher level of education, and household heads were more often
in services or business occupations.
Education was found to
be directly related to the use of preventive health services
(Mechanic, 1972).
In Nigeria, Akesode (1982:312) observed
that the education level of parents had a direct effect on
attendance at infant welfare clinics and hence on the
immunization status of children.
In another study of
childhood immunization in Indonesia, Singarimbun et al.,
(1986:9) also observed that maternal education and higher
economic status were associated with a higher degree of
completeness of immunization.
Similar results were also
obtained in Malaysia by Ramlah (1986) where maternal education
was found to be associated with a higher immunization status
of the children.
She argued that the higher proportion of
immunized children among the better educated mothers reflects
their greater understanding of the benefits of immunizations
and also their ability to maximise the services offered by the
Government.
Objective of the Study
The present study aims to investigate the levels of
overall knowledge of immunizations and attitudes towards
immunization as a means of prevention of immunizable diseases.
It will also try to identify factors associated with the
acceptance of childhood immunizations.
The Study Area
The data for this study were collected in ten villages in
Matlab upazila (administrative unit below a district),
Bangladesh.
The International Centre for Diarrhoeal Diseases
Research, Bangladesh (ICDDR.B, formerly the Cholera Research
Laboratory - CRL) has been operating a Demographic
As of
Surveillance System (DSS) in this area since 1966.
July 1983, the DSS area had 191,006 people of whom 96,451 and
94,555 lived in the Intervention area and the Comparison area
respectively (Shaikh et al., 1985:2).
The Demographic
Surveillance System includes regular cross-sectional censuses
and continuous registration of births, deaths, migration,
Other activities of the ICDDR.B in
marriages and divorces.
the study area include a diarrhoeal treatment program, family
planning and other health interventions.
During 1978 ICDDR.B
introduced village-based MCH-FP (Maternal-Child Health and
Family Planning) services in the intervention area.
Later
tetanus-toxoid immunization of pregnant mothers and more
recently a Measles immunization program have also been
introduced in the Intervention area.
On the other hand the
activities of the ICDDR.B health workers in the Comparison
area are mainly limited to registration of vital events and
providing referral for diarrhoeal treatment in Matlab centre.
More details of the field operation have been reported
elsewhere (see CRL, 1978) .
The Matlab field research area is on the low-lying
deltaic plain of Bangladesh, situated at a distance of about
45 kilometers south-east of the capital, Dhaka.
A tidal
2
river and its several canals flow through the area and are
used as the major transportation routes within and outside the
area.
The area is relatively dry during the months of
December to February with very low rainfall.
May to
September is the monsoon season when the water table rises and
inundates most of the agricultural and non-residential area.
A motorable road links the upazila headquarters with the
district headquarters at Chandpur, but communication within
the area is only possible on foot or by boat.
As in the greater part of Bangladesh, farming is the
dominant occupation of the heads of households.
Except for a
few villages where fishing is the main sources of livelihood,
agriculture is the principal economic activity of the area.
At Matlab upazila health centre immunization services for
DPT (vaccine against diphtheria, pertussis and tetanus), Polio
(vaccine against polio), Measles (vaccine against measles),
BCG (vaccine against tuberculosis) and DT (vaccine against
diphtheria and tetanus) are available as a part of the
national EPI program.
The uni-purpose EPI technician is
responsible for the delivery of immunization services once a
week at the Matlab center and once a month at four outreach
bases, covering about two hundred thousand population in the
upazila.
In the Intervention area ICDDR.B introduced a
Measles immunization program during 1982 as a part of maternal
and child health and family planning activities in that area.
Methodology and Sampling
The data for this study were collected by a KAP study of
childhood immunization in Matlab DSS area which was conducted
during January-March 1986.
The survey was carried out under
the auspices of the Child Survival Project, IPDP, The
Australian National University and the author was the
principal investigator of the survey.
Owing to time
constraints and logistic difficulties in the fieldwork, the
survey was restricted to a sample of 10 villages selected from
33 villages within approximately three miles of the Matlab
health centre.
Half the villages were selected randomly from
the Intervention area and the remaining half from the
Comparison area.
The rationale behind this was to facilitate
a comparison between the Intervention and the Comparison
areas.
A sample of 330 currently married women (30 per cent of
the eligible respondents), who had at least one living child
under the age of five years, was randomly selected from each
of the selected villages.
The rationale of selecting an
equal number of respondents from each village was to ensure a
sufficient number of respondents for each village.
Larger
villages have more than one FVW's (Female Village Worker)
working area.
To minimise the possible bias due to different
village sizes, only one FVW's working area was selected from
each of the larger villages.
As the objective of the study
was not to estimate the coverage of immunizations precisely
but rather to examine the effect of social and demographic and
cultural factors on immunization practice, probability
3
In the
proportional to sizes sampling was not employed.
analysis Z-test statistic for the difference of two
proportions and Chi-square statistic for the test of equality
of more than two proportions will be used.
Data Collection
The survey was carried out by three female interviewers,
who were hired locally in Matlab.
All were married, aged
between 20 and 25, and all of them had a college education.
All had extensive experience with the local village life, and
had previously worked on projects undertaken by ICDDR.B.
Of the 330 selected, 304 eligible respondents were
successfully interviewed and the response rate was 92 per
cent.
In some cases two visits were made for a successful
interview.
The information on the practice of childhood immunization
was retrospective.
Although the Bangladesh Government
recently launched the EPI program with the aim of increasing
immunization coverage to 85 per cent of target group, children
under the age of two years, by the end of 1990, the actual
coverage throughout the nation is very low.
This resulted in
very few respondents in the sample having had their children
immunized.
This caused difficulties in the study of
differentials in the practice of immunizations.
Interviewers
were instructed to look for scar signs on the arm to check
whether the child had received BCG, to ask respondents whether
their child had received DPT in the leg and Polio orally.
However, there is still the chance of misreporting of one
vaccine for the other or any other injections (non-vaccine).
While asking about knowledge of immunizations, probing
with local terms, which incorporate the names of the diseases,
was used for BCG and Measles.
This may overestimate the
level of knowledge of the function of these immunizations.
Memory lapse may also result in misreporting of the age of the
child when he/she was immunized.
Knowledge of Immunization
To assess the general knowledge of immunization, the
respondents were asked whether they had heard about
immunization in general.
While a high proportion of
respondents in both the Intervention area, 81.8 per cent, and
the Comparison area, 74.7 per cent, had heard of immunization
in general, either as pratisedhak tika (meaning which makes
one invulnerable against diseases), or foot or injection, the
knowledge of specific immunizations was quite low.
Very few
respondents had heard the names of the vaccines BCG, Measles,
DPT and Polio.
However, when the question was asked with the
term zakshar tika (meaning injection against tuberculosis) and
luntir tika (meaning injection against measles) instead of
"BCG” and "Measles" respectively, quite a large number of
respondents mentioned that they had heard about these
vaccines.
The highest percentage of respondents in the total
area who had ever heard of the specific immunization was for
4
Measles, 52.6 per cent; followed by BCG, 48.4 per cent; DPT,
7.2 per cent; and Polio, 2.3 per cent.
Knowledge about the Function of Immunization
Knowledge about which diseases are prevented by specific
immunization may be associated with the mother's motivation in
seeking specific immunizations for their children.
When the
knowledge of the respondents about what diseases these
immunizations (DPT, BCG, Polio and Measles) protect against
was examined, it was found that only very low proportions of
respondents knew the correct function of most of these
immunizations.
There were usually a number of local names
for diseases, for example, measles was known as lunti and ham,
whooping cough as kui-kash or meadi-kash (meaning long
duration cough), tuberculosis as zaksha and tetanus as takuria
or alga or dhanustanker.
In the Intervention area the
highest proportion of respondents knew the correct function of
Measles vaccine, 64.9 per cent, followed by BCG, 36.4 per cent
(Table 1).
The level of knowledge of the other two
immunizations, DPT and Polio were negligible, with only 3.2
per cent and 1.3 per cent of the respondents having correct
knowledge about DPT and Polio immunizations respectively.
It
is not surprising that in the Intervention area the level of
correct knowledge about Measles vaccine was quite high
compared to other immunizations.
In the Intervention area
ICDDR.B introduced a Measles immunization program during 1982
covering half of the area, and later in 1985 the program
In
coverage was extended to the remaining half of the area,
the Intervention area Measles immunization services are
delivered at home through village-based FVWs.
The second highest proportion of respondents, 36.4 per
cent, knew the correct function of BCG.
When asking about
immunizations probing with the local term zakshar tika was
That might have overestimated the level
used instead of BCG.
of knowledge of BCG as the term used incorporates the name of
the disease.
In the Comparison area, the highest level of knowledge of
the function of immunization was for BCG followed by Measles
Fifty-two per cent of respondents knew the
(Table 1).
correct function of BCG and quite a high proportion, 30.7 per
cent, knew what diseases Measles immunization protects
against.
Although the ICDDR.B Measles immunization program
did not include the Comparison area, it is not unlikely that
some respondents had access to Measles immunization services
while visiting their relatives in the neighbouring
intervention area, and that could be one reason why a
comparatively high proportion of respondents knew the correct
function.
In both the Intervention and Comparison areas very few
respondents correctly identified the function of Polio and DPT
vaccines.
For DPT all the respondents who knew the correct
function mentioned only diphtheria, and no respondent
mentioned that it protects against whooping cough and tetanus.
5
Table 1
Percentage of Respondents Who Knew the Function of DPT, BCG,
Polio and Measles Immunization
Knowledge of function
Type of Vaccine
BCG
Polio
DPT
Measles
Intervention Area
Protect against
correct diseases
3.2
36.4
1.3
64.9
Protect against
incorrect diseases
or don't know
4.5
7.2
1.3
7.8
Never heard of it
92.2
56.5
97.4
27.2
Total
N
99.9
100.0
100.0
99.9
(154)
Comparison Area
Protect against
correct diseases
4.0
52.0
0.7
30.7
Protect against
incorrect diseases
or don't know
2.7
1.3
2.3
0.7
Never heard of it
93.3
46.6
97.0
68.5
Total
N
100.0
99.9
100.0
99.9
Notes:
(150)
1. In the case of DPT all the respondents who knew the
function mentioned only diphtheria.
2. Total percentage not always 100.0% because of
rounding.
Perception of Disease Seriousness
The perception of the severity of illness caused by these
immunizable diseases may influence immunization decision
processes for the mothers.
It is important to examine
whether the respondents consider the immunizable diseases as
dangerous to the children.
The results indicate that at
least three of these immunizable diseases were considered
dangerous (potentially fatal) by the respondents (Table 2).
In both the Intervention and the Comparison areas there
was a clear pattern of measles and tuberculosis being the
major dangerous diseases, followed by whooping cough.
In the
Intervention area 77.9 per cent of the respondents considered
6
measles as dangerous to the children.
Quite a high
proportion of respondents also considered tuberculosis, 48.7
per cent, and whooping cough, 40.9 per cent, as potentially
dangerous to the children.
Very few respondents mentioned
diphtheria, tetanus and polio as potentially dangerous.
In the Comparison area a similar pattern of perception of
disease seriousness was observed.
However, a lower
proportion of respondents mentioned each of these diseases as
dangerous.
It seems that respondents in the Comparison area
were less aware of the seriousness of those diseases.
Table 2
Percentage of Respondents Considering Specific Diseases
as Potentially Dangerous to the Health of Their Children
Diseases
Intervention Area
Comparison Area
Measles
Tuberculosis
Whooping cough
Diphtheria
Tetanus
Polio_________
N_____________
77.9
48.7
40.9
8.4
9.1
5.8
(1541
60.7
40.0
20.0
2.0
2.0
0.8
■(15 QI
Note:
Respondents could mention more than one disease.
The observed pattern of disease seriousness was not
consistent with the ranking of the known world-wide case
fatality (CF) rates of these diseases, as tetanus,
tuberculosis and whooping cough have CF rates greater than 25
per cent while measles, diphtheria and polio have rates below
10 per cent (Benenson, 1985:233).
As mentioned previously,
tetanus has a few local names.
Dhanustanker in rural areas
of Bangladesh is considered a disease of adults, while alga
This
and takuria are considered as diseases during infancy,
definitional difference could be the reason that only a few
respondents mentioned tetanus as dangerous to children, even
though it is the major cause of infant and child deaths in the
The reason that very few respondents mentioned
locality.
polio and diphtheria as dangerous could be that they were
ignorant about these diseases.
The prevalence of diphtheria
and polio was also low.
However, the prevalence rates were
not available.
Measles and whooping cough were fairly common
diseases in the locality and that could be the reason why a
large proportion of respondents considered those diseases as
dangerous.
Attitudes and Beliefs about the Prevention of Diseases
It is important to examine mothers' beliefs and attitudes
towards the prevention of diseases which may be associated
with the practice of immunization.
Table 3 shows the
7
proportion of respondents suggesting specific methods of
prevention for specific diseases.
The largest proportion of
respondents responding to the question was for measles,
followed by tuberculosis.
Though whooping cough was a fairly
common disease in the locality, very few respondents suggested
any method for the prevention of whooping cough.
Diphtheria
has a low prevalence and a negligible proportion of
respondents suggested any method of prevention of that
disease.
The results show that immunization was not considered as
the only means to prevent certain diseases.
There were local
beliefs and local customs for the prevention and treatment of
diseases.
In the rural area of Bangladesh many believe that
some diseases can be cured by daktari cikitsay (allopathic
treatment), some by kabiraj (indigenous healers) and some by
faith healing, for example, by using a tabiz (amulet) or with
In the
panipura (consecrated water) (Malony et al., 1981:7).
Intervention area, while 56.4 suggested a method for
prevention, only 30.5 per cent suggested that immunization or
injection could prevent measles.
Because the term
"injection" might be confused with the other injections (non
vaccine injections), the actual proportion of respondents
recommending immunization might be even lower.
Eight per
cent of respondents suggested seeing the faith healer or
staying ritually clean and observing certain rules and
customs, for instance, wearing sanctified amulets.
A large
proportion of respondents, 13.6 per cent, who responded to the
question could not specifically mention any method, but
suggested that one need to be careful and not to make contact
with the infected person.
The same pattern was observed for prevention of
tuberculosis and whooping cough, where less than half of the
respondents who suggested any method, recommended immunization
as the means to prevent these diseases (Table 3).
However,
only a few respondents, 4.7 per cent, suggested any method for
the prevention of diphtheria and more than half of these
suggested immunization as a preventive measure.
The reason
could be that most of those respondents who suggested any
method for diphtheria had higher education and were
economically well off.
In the Comparison area, immunization as a preventive
measure received even less importance.
While 52.6 per cent
of respondents suggested some method for prevention of
measles, only 6.7 per cent recommended immunization.
The
major proportion of respondents could not mention any specific
method other than suggesting that one needed to be careful
about the disease and not to make any contact with the
infected person.
Seeking advice from the faith healer and
being ritually clean appears to be in second place (Table 3).
A similar pattern was observed for tuberculosis, whooping
cough and diphtheria, where a very low percentage of
respondents perceived immunization as a means to prevent these
diseases.
For tuberculosis and whooping cough less than one
8
Table 3
Percentage of Respondents Recommending Specific Methods/
Suggestions for the Prevention of Specific Diseases
Suggested method
Diphth
TB
Disease______
Measles
WC
Intervention Area
Immunization/
Injection
2.6
11.0
30.5
9.0
Spiritual healer or
rituals
0.7
3.2
8.4
2.6
No contact with
infected/Careful
0.7
10.4
13.6
9.0
Others
0.7
2.7
3.9
3.9
72.7 _____ 43.6
100.0
100.0
(154)
75.5
100.0
Don't know/Never
heard of this disease
Total
N
95.3
100.0
Comparison Area
Immunization/
Inj ection
0.7
9.3
6.7
4.0
Spiritual healer or
rituals
0.0
0.0
11.3
4.0
No contact with
infected/Careful
0.7
15.3
26.6
8.0
Others
0.0
4.7
8.0
0.7
70.7 _____ 47.4
100.0
100.0
(150)
83.3
100.0
Don't know/Never
heard of this disease
Total
N
Notes:
98.4
100.0
1. "Others" includes suggestions by husband, relatives
and village doctors.
2. For the prevention of polio only four respondents
responded to the question.
Three recommended
immunization and the other one mentioned that the
husband would decide what to do.
3. Diphth - diphtheria, TB - tuberculosis,
WC - whooping cough.
third of the respondents who could specify any method
suggested immunization.
Very few respondents answered for
diphtheria, and the results do not show any pattern.
9
Practice of Immunization
This section discusses the practice of four major types
of immunization, DPT, BCG, Polio and Measles.
The levels and
patterns of use of immunization services will be examined by a
number of variables, including level of correct knowledge,
socio-economic, and demographic characteristics of mothers.
In Bangladesh immunization coverages for all four
vaccines are still quite low.
Although the Bangladesh
Government has recently undertaken an ambitious program of
immunization to achieve the coverage target of 85 per cent of
children under age 2 years by the year of 1990, immunization
coverage at the time of this study was below 2 per cent for
all four major vaccines, i.e. DPT, BCG, Polio and Measles.
Results from the Matlab study population also show that except
for Measles immunization coverages for DPT, BCG and Polio were
very low (Table 4).
In the study area 23 per cent of the
respondents had their youngest child immunized with Measles
followed by BCG, 3.3 per cent, DPT, 3.0 per cent, and Polio,
1.3 per cent.
In contrast to Measles, the coverage rates for
the other three immunizations, DPT, BCG and Polio, do not show
any significant differences between the Intervention and the
Comparison areas (Table 4).
The reason for the considerably
higher coverage of Measles in the Intervention area, 39.6 per
cent, as compared to the Comparison area, 6.0 per cent, could
be due to the Measles immunization program introduced by
ICDDR.B in the Intervention area.
As immunization coverages for DPT and Polio were very
low, in the analysis a child will be considered immunized with
DPT and Polio if they had received at least one dose of DPT
and Polio.
Table 4
Percentage of Respondents Who Had Their Youngest Child
Immunized with Specific Immunizations
Immunizations
Intervention
Area
Comparison
Area
3.3
Total
2.3
0.7
DPT (3 doses)
(1 or 2 doses)
1.9
0.7
BCG
3.2
3.3
3.3
Measles**
39.6
6.0
23.0
Polio
N
1.9
(154)
0.7
(150)
1.3
(304)
Note:
2.6
2.6
0.7
* p<0.05, ** p<0.01 (based on Z-test).
10
3.0
Immunization Practice by Knowledge of Diseases Prevented
It could be argued that knowledge about what diseases are
prevented by specific immunization might be associated with
the acceptance of immunization.
Several studies found that
mothers with correct knowledge about the function of the
specific immunization were more likely to have their children
immunized than mothers with incorrect or no knowledge
(Singarimbun, 1986:7; Markland, 1976:168).
In this study a
consistent pattern of higher coverages of the specific
immunizations was also found to be associated with higher
levels of knowledge of the correct function of each of the
four types of immunization.
In the Intervention area the coverage rates of DPT, Polio
and Measles immunizations were significantly higher for
mothers with correct knowledge of the function of the
immunization than for mothers with incorrect knowledge or no
knowledge (Table 5).
In the case of Measles 56.0 per cent of
the respondents who knew the correct function of that
immunization had their child immunized with Measles, while
only 9.2 per cent of the respondents who did not know the
function correctly had their child immunized with Measles.
In the case of DPT it increased from zero per cent to 80 per
cent and for Polio 1.3 per cent to 50 per cent.
For BCG the
difference in the coverage rates between mothers with correct
Table 5
Percentage of Respondents Whose Youngest Child Had Been
Immunized with Specific Immunizations According to
Knowledge of Correct Diseases Prevented
Type of vaccine
Intervention Area
Incorrect
Correct
knowledge knowledge
Comparison Area
Correct
Incorrect
knowledge knowledge
DPT
(80.0)**
(5)
0.0
(149)
(67.0)**
(6)
0.7
(144)
BCG
7.1*
(56)
0.5
(98)
6.4*
(78)
0.0
(72)
Polio
(50.0)**
(2)
1.3
(152)
(100.0)**
(1)
0.0
(149)
Measles
56.0**
(100)
(1541
9.2
(54)
N
Note:
17.4**
1.0
(104)
(46)
£1501
For each type of vaccine the numbers in parentheses
represent the total number of respondents who had
correct knowledge/incorrect knowledge.
* p<0.05, ** p<0.01 (based on Z-test).
11
knowledge and mothers with incorrect knowledge was also
significant, and the observed pattern showed higher acceptance
among mothers with correct knowledge.
In the Comparison area, the pattern of the proportions of
children receiving any vaccination by knowledge of function
was similar to that in the Intervention area.
For each of
the four types of immunization the coverage rates were higher
for mothers having correct knowledge.
Measles coverage rate
rose from one per cent for respondents with incorrect or no
knowledge to 17.4 per cent for respondents with correct
knowledge.
For DPT it rose from 0.7 per cent to 67.0 per
cent, for BCG from zero per cent to 6.4 per cent, and for
Polio the only person who knew the correct function had her
child immunized with that vaccine.
Although levels of use
are low, the pattern of positive association between knowledge
and use is clear.
There were also strong positive associations between use
of immunization and correct knowledge of appropriate age
ranges (Chakraborty, 1987:37), of appropriate number of doses
(ibid, p.38), and of belief that the relevant disease is
dangerous (ibid, p.39).
There were also considerably higher levels of
immunization for the four major vaccines where both parents
had some education (between 17 and 29 percent) compared to
those where both parents were uneducated (0 to 12 percent).
This pattern held for all vaccines, though measles coverage
was much higher even where only one parent, usually the
husband, was educated (ibid, p.44).
The demographic
characteristics of maternal age, and parity of the child did
not show any strong association with immunization coverage
(ibid, p.46).
Reasons for Incomplete or No Immunization
Respondents were asked why their youngest child had not
been immunized with any of the available immunizations.
As
immunization coverages were very low for all immunizations
except Measles, the analysis will not examine the reasons for
non-immunization for each of the immunizations separately.
The major reason given for not obtaining any
immunizations, both the Intervention and in the Comparison
areas, was that the immunization worker did not come to the
house of the respondent.
This passive attitude to the
delivery of a preventive health intervention was also noted in
Java by Streatfield and Singarimbun, 1988:1242).
The lack of
services is a problem in a society where women cannot easily
travel to the clinic or hospital with their child.
Of those
who had ever heard of immunizations but never obtained any
immunizations, 56 per cent in the Intervention area and 64 per
cent in the Comparison area gave that reason.
As Measles
immunization was offered at home by ICDDR.B in the
Intervention area, it is not surprising that the major
proportion of respondents expect immunization services at home
12
in the Intervention area.
Other immunizations are only
available at the Matlab upazila health centre.
The second major reason in both the areas was that "child
was too young to immunize".
The EPI recommend immunization
for DPT and Polio from six weeks and BCG from birth up to two
years of age and for Measles nine months to two years.
Although at the time of the survey the age of the youngest
child of all respondents was a minimum of three months, a high
proportion of respondents, 21.0 per cent in the Intervention
area and 11.5 per cent in the Comparison area, gave the reason
that the child was too young (Table 6).
It seems that they
were ignorant about the correct age of immunization.
Table 6
Percentage of Respondents Mentioning Specific Reasons for
Not Having Youngest Child Immunized with Any Immunizations
Reasons
Intervention Area
Comparison Area
Immunization worker didn't
come to house
56.5
64.6
Child was too young
21.0
11.5
Child wasn't at home when
immunization worker came
11.3
2.1
Forbidden by husband or
father/mother-in-law
6.5
3.1
Child cried and didn't
like immunizations
1.6
5.2
Didn't like to go to
hospital for immunization
0.0
8.3
Others
N_____
6.5
(62)
£961
Note:
5.2
"Others" include "child was sick when immunization
worker called", "heard that child get sick after
immunization" and "no specific reason".
The third reason for non-immunization in the Intervention
area was that the child was not home when the health worker
called, 11.3 per cent mentioning that reason.
This is
surprising because in the Intervention area at least Measles
vaccine was easily available from village based FVWs.
In the
Comparison area the third major reason was that the respondent
did not like to go to hospital to receive the immunization and
it seems that they do not consider immunizations to be
valuable enough to spend time and effort to get them from the
Matlab hospital.
However, when they were asked whether they
13
would accept immunization if provided at home, many said that
they would.
Among the other reasons for not obtaining an immunization
were that respondents were "forbidden by their husband or
mother-in-laws", "children cried at the time of immunization",
"child was sick", and "did not think immunizations are
effective measures against immunizable diseases".
Conclusion
To summarize the patterns associated with immunization
coverage , the overall knowledge of immunization in both the
Intervention area and the Comparison area was quite low.
Although for Measles and BCG vaccines, quite a high proportion
of respondents had correct knowledge of the function, very few
had correct knowledge of the appropriate age range and
required number of doses.
Immunization coverage for all vaccines were very low,
except for Measles which was comparatively higher in the
Intervention area due to the ICDDR,B Measles immunization
program.
Acceptance of immunization services for most of the
vaccines was found to be positively correlated with overall
knowledge of immunization.
Perception of disease seriousness
had a positive relation with acceptance of immunization
against those diseases, except for tuberculosis.
These findings indicate that if Bangladesh is going to
achieve the higher levels of child immunization, as part of
its EPI objectives, then there is a need for more effective
education of the public concerning the potentially serious
consequences of a number of immunizable diseases, and the
disease preventing function of immunization.
In regard to the delivery of immunization services, the
improved coverage of Measles vaccine due to the home visit
program tested in the Matlab Intervention area strongly
suggests that coverage of other vaccines will be increased if
the service can reach out from the clinics to delivery points
closer to the homes of the target children.
Such an approach
is particularly important in a society where it is not always
socially or culturally acceptable for mothers with small
children to travel unaccompanied to a clinic or hospital.
14
References
Akesode, F.A., 1982. "Factors Affecting the Use of Primary
Health Care Clinics for Children", Journal of
Epidemiology and Community Health, 36:310-314.
Chakraborty, Nitai, 1987. "Factors Affecting Childhood
Immunization in Rural Bangladesh", unpublished thesis for
the degree of MA in Demography at the Australian National
University, Canberra.
Chen, L.C., M. Rahaman and A.M. Sarder, 1980. "Epidemiology
and the Cause of Death Among Children in a Rural Area of
Bangladesh", International Journal of Epidemiology,
9(1):25-33.
CRL (Cholera Research Laboratory), 1978. Demographic
Surveillance System - Matlab, Vol.1. Dhaka: International
Centre for Diarrhoeal Disease Research, Bangladesh.
Government of Bangladesh, 1985a. Expanded Programme on
Immunization: A Draft Project Document Covering the
Period July 1985-June 1986. Dhaka: Office of the Director
General of Health Services, Ministry of Health,
Government of Bangladesh.
, 1985b. Expanded Programme on
Immunization. Unpublished Service Statistics for 1984-85.
Office of the Director General of Health Services,
Ministry of Health, Government of Bangladesh.
Malony, E. et al., 1981. Beliefs and Fertility in Bangladesh.
Dhaka: International Centre for Diarrhoeal Disease
Research, Bangladesh.
Markland, R.E. and D.E. Durand, 1976. "An Investigation of
Socio-psychological Factors Affecting Infant
Immunization", American Journal of Public Health,
66(2):168-170.
Mechanic, D., 1972. "Sociology and Public Health Perspectives
for Application", American Journal of Public Health,
62:146-149.
Population Information Program, 1986. "Issues in World Health,
Population Reports, Series L., No.5.
Rahaman, M. et al., 1986. "Immunization Acceptance Among
Pregnant Women in Rural Bangladesh", Bulletin of The
World Health Organization, 60(2):269-277.
Ramlah, H.M., 1986. "Education and Preventive Health
Services", Child Survival Research Note, No.6, Child
Survival Project, International Population Dynamics
Program, Demography Dept., The Australian National
University.
15
” S’
Shaikh, et al., 1985. Demographic Surveillance System Matlab, Vol.14, Scientific Report No.64. Dhaka:
International Centre for Diarrhoeal Disease Research,
Bangladesh.
Singarimbun, M.r K. Streatfield and I. Singarimbun, 1986.
"Some Factors Affecting the Use of Childhood
Immunization”, Child Survival Research Note, No.7, Child
Survival Project, International Population Dynamics
Program, Demography Dept., The Australian National
University.
Streatfield, K. and M. Singarimbun, 1988. ’’Social Factors
Affecting Use of Immunization in Indonesia", Social
Science and Medicine, 27(11):1237-1245.
V
5
16
F.
Research Note on
CHILD
SURVIVAL
31CS
Number
Date
15 November 1990
International Population Dynamics Program
Department of Demography
Research School of Social Sciences
The Australian National University
Canberra, ACT, Australia
A Project of The Department of Demography
The Australian National University
Sponsored by The Ford Foundation
DEMOGRAPHIC, SOCIOECONOMIC AND HEALTH-RELATED EFFECTS ON
COMPONENTS OF PHILIPPINE CHILD MORTALITY
Josefina V. Cabigon
Demography Program,
Division of Demography and Sociology,
Research School of Social Sciences,
The Australian National University
Note:
•V
•ANO
• N F O F M A T10 N
CEMf £
Child Survival Research Notes are brief discussions
of issues of current relevance to researchers and
policy-makers concerned with problems of high infant
and child mortality in the world. The International
Population Dynamics Program, Department of Demography,
The Australian National University, distributes these
notes with their regular Bibliographic Circular.
Production of the Child Survival Research Notes is
made possible through a grant from the Ford Foundation.
I Responsibility for the content of Child Survival
|Research Notes rests with the author(s) alone, and not
the above-listed organisations.
Introduction
Several issues have drawn attention to a more comprehensive analysis of covariates of Philippine child
mortality. The first is the call of planners for more substantive and scientific studies to supplement
operational researches in providing insights for the determination of the best mix of strategies for
preventing child deaths within the constraint of available health resources and manpower. A response of
the Philippine government to the global promotion of primary health care and child survival led by
international agencies like the World Health Organization (WHO) and the United Nations Intemationi
Children’s Emergency Fund (UNICEF) necessitates identification of the underlying causes of child
mortality.
The second issue is the consistent indication that the long-term trend in Philippine child mortality decline
has plateaued since 1960 (Cabigon 1990; Conccepcion and Smith 1977; Madigan 1977; Reyes 1981;
United Nations 1982; Zablan 1978, 1983). The emergence of more detailed and relevant Philippine data
and of sophisticated procedures for dealing with qualitative data in the analysis of determinants of child
mortality provides a basis for in-depth analysis yielding some insights into such a trend.
The third issue deals with the part played by demographic or health-risk factors: maternal age (age of
mother at childbirth), parity (birth order), pace of childbearing when they are systematically studied with
other variables. Literature has been consistent in demonstrating the powerful effects of maternal age and
birth order on infant and child mortality, but the effects of child-spacing have been the subject of further
investigation due to the methodological difficulties and unclear types of mechanisms that operate (Ballweg
and Pagtolun-an 1988; Boulier and Paqueo 1988; Casterline, Cooksey and Ismail 1989; Cleland and Sathar
1984; Cramer 1987; DaVanzo, Butz and Habicht 1983; De Sweemer 1984; Edouard 1981; Frenzen and
Hogan 1982; Hobcraft, McDonald and Rutstein 1983; Hull and Gubhaju 1986; Kiely, Paneth and Susser
1986; Knodel and Hermalin 1984; Martin, Trussell, Salvail and Shah 1983; PaUoni 1989; Palloni and
Tienda 1986; Rutstein 1983; Trussell and Hammerslough 1983; Yudkin and Baras 1983). With the current
Philippine national population program emphasizing the health benefits of family planning, the role of the
health risk factors on child mortality needs to be further studied. What will happen with the role of birth
order and maternal age at childbirth when other demographic, socioeconomic and health-related variables
are taken into account? Is there a difference in the patterns of their net effects on various ages of child
mortality?
The fourth issue pertains to the effects of education and household income on child mortality. While
parental education (mother, father or both) strongly influences child mortality in the Philippines and in
other countries (Soulier and Paqueo 1988; Cabigon 1982; Caldwell 1979; Caldwell and McDonald 1981;
Concepcion 1982; Concepcion and Cabigon, 1982; Cramer 1987; Haines and Avery 1982; Hobcraft et al.
1984; Hull and Gubhaju 1986; Madigan n.d.; Martin et al. 1983; Trussell and Hammerslough 1983;
Victora, Smith and Vaughan 1986), explanations of this relationship appear to be complicated. While there
are reasons for arguing that the higher the educational attainment of parents, the lower the infant and child
mortality, there is some uncertainty whether the measured effects are attributable to schooling in itself or to
other characteristics such as economic status or various social roles adopted by people with different levels
of education (Hull and Gubhaju 1986, pp. 116-117).
In addition. Hobcraft et al. (1984, pp. 219-220) noted the difficulty of interpreting socioeconomic
differences in infant and child mortality, as the five socioeconomic factors they considered - mother’s
education, mother’s work status, husband’s occupation, husband’s education and mother’s place of
residence at survey date - are all more or less remote in the causal chain leading to prevented deaths in
early life. They contended that there are mechanisms other than formal schooling, which were not included
in their model and may have been operating, such as income, mother’s work habits, supply of medical and
health care, and public health provisions, such as water supply, sewerage and refuse disposal. They also
argued that education may succeed in ending traditional and unhealthy practices, although this may not
require formal schooling.
2
Drawing on household production theory, Schultz (1980) also hypothesized that better educated women
earn more in the labour market and marry better educated men; consequently they have higher family
incomes enabling them to purchase goods and services that improve child health. Education may also
increase the effectiveness of women’s non-market child-care activities although as Schultz cautioned, the
fact that paid work requires women to be absent from the home can have an offsetting negative effect on
the quality of child care.
Furthermore, in rural northern Thailand, Frenzen and Hogan (1982) found no support for Caldwell’s
(1979) findings that maternal education and parental beliefs about wealth transfers are crucial factors
affecting infant mortality declines in developing societies. These two variables become insignificant after
adjusting for health information, social class, district development level and whether births are wanted.
One analytical strategy to shed some light on these uncertainties is to examine interaction or the extent to
which the effect of education on child mortality varies according to various categories of other important
variables. Studies found household income or its proxy, poverty, to be strongly associated with infant or
child mortality (Casterline et al. 1989; Gortmaker 1979; Madigan n.d.; Victora et al. 1986). It is then
worthwhile examining whether household income plays an important role in Philippine child mortality. If
so, does the relationship between maternal education and child mortality then differ by income levels?
Conversely, is the relationship between income and child mortality different for the low and highly
educated groups? Moreover, because of the unavailability of reliable data on income in the Philippines,
education has been taken as a proxy of socioeconomic status. If income is more important than education
as a determinant of child mortality, then education is not an adequate measure of socioeconomic status in
the Philippines.
The fifth issue relates to the impact of environmental factors on child mortality. Access to piped water
plays an important role in early childhood in Brazil (Merrick 1985; Victora et al. 1986) and on infant
mortality in Egypt (Casterline et al. 1989). Toilet facility is associated with lower child mortality in Sri
Lanka1 (Trussell and Hammerslough 1983), and with post-neonatal mortality in Malaysia (DaVanzo et al.
1983), but has no influence on infant and child mortality in Egypt (Casterline et al. 1989). Housing quality
affects child mortality in Costa Rica (Haines and Avery 1982) and in a rural Philippine province (Johnson
and Nelson 1984). Public sanitation and provision of safe drinking water are health programs of the
Philippines. Assessing their roles in influencing child mortality is useful in health planning.
The last issue pertains to methodological problems. Long (1987, pp. 3-7) drew attention to common
mistakes in quantitative social research resulting mainly from the rapid growth in the variety and
complexity of methods available. These errors are both failure to apply appropriate techniques and
inappropriate application of statistical methods. He stated that despite the frequent occurrence of these
errors which often have solutions in the statistical literature, there are few articles or books that provide
clear, practical, and accurate discussions of the issues.
On the use of weights in analysing survey data, Lee, Forthofer and Lorimor (1986) highlighted the
problems and strategies for analysing survey data from complex sample designs. They presented three
examples which indicate that sample weights are important in estimation, and standard errors, calculated by
assuming simple random sampling (even with use of weights), do not always agree with those obtained by
more appropriate methods that take the sample designs into account. Moreover, statistical packages do not
always handle weighted data adequately.
In the statistical literature, there are arguments for and against the use of sampling weights. For instance,
Clogg and Eliason (1987, pp. 21-27) and T. W. Pullum (personal communication. May 1, 1989) stated that
in data sets weighted on a case-by-case basis, it is incorrect strategy to analyse the unweighted data by
simply ignoring the weighting feature altogether, or analyse weighted data as if they were obtained from a
data set without any weighting features. Unweighted estimated parameters are clearly biased and their
corresponding standard errors are incorrect Use of weights, which has been the common approach2, will
yield unbiased parameter estimates, but estimates of the standard errors and all other statistical tests will be
biased. Lee et al. (1986) and Clogg and Eliason (1987) illustrated these biases clearly and suggested some
solutions.
This paper presents insights into the above issues, through an exhaustive exploratory analysis and
simultaneous examination of several demographic, socioeconomic and health-related variables to identify
those which have a net influence on neonatal, post-neonatal, infant and child mortality and test for
significant interactions between or among the emerging important variables. It uses appropriate
methodological tools and strategies for analysing survey data. In the Philippine studies reviewed above,
there has been no systematic study of the extent to which the effects of important variables may have been
greater or less for various subgroups of the population in question. Interaction between or among the
3
variables so far found to influence Philippine child mortality may be important; therefore it needs to be
assessed both for policy significance and theoretical interest. The next section further elaborates these
points.
Methods
Data
The analysis was based on the 1983 National Demographic Survey (NDS), conducted by a consortium of
research centres composed of the University of the Philippines Population Institute (UPPI), University of
San Carlos Office of Population Studies (USC-OPS), and Xavier University Mindanao Centre for
Population Studies (XU-MCPS). The purpose of the survey was to gather information on fertility,
mortality, migration, nuptiality, labour force participation and family planning.
The 1983 NDS collected information from a nationally representative sample of 13,000 households. To
allow for regional-level analysis in line with the regional thrust of the government, a sample of 1,000
households per region was obtained.
The sampling design featured a two-stage sampling scheme where the primary sampling unit, barangay2
was drawn with replacement and with probability proportional to the number of households per barangay,
in each region. The ultimate sampling unit was the household and was drawn systematically with a random
start.
Relevant child mortality data used in this article originated from the pregnancy history.4 Background
variables, such as socioeconomic, health-related and other demographic characteristics were drawn from
the household and ever-married women files.
Births to ever-married women were units of observation, totalling 42,471, reconstructed from the
pregnancy history record. The following were the factors examined for each birth.
Demographic factors', (a) birth order (first, second and third, and fourth and higher); (b) maternal age at
birth of child (in years: less than 20, 20-34, and 35 and over); and (c) length of preceding birth interval (in
months: less than 18, 18-30, and 31 and over). The present study focused on the estimation of the the
effects of the first two factors. The most recent work of Palloni (1989) uncovered the persistent
generalized and strong effects of the preceding and the following interval even when controlling for
contextual variables such as mortality levels, breastfeeding patterns and contraceptive prevalence and for
individual variables, such as mother’s education and access to information. Palloni also found little support
for the hypothesis that early cessation of breastfeeding is the main mechanism through which the negative
effects of birth interval operate. Hence, it was important to allow for the role of the preceding birth interval
in the present analysis. Nevertheless, because of the methodological problems and complicated mechanism
through which it operates in the causal chain leading to deaths in childhood, preceding birth interval was
treated as a control.5
Socioeconomic factors: (a) current residence (rural and urban); (b) presence of electricity in the household
(with and without); (c) average household monthly income (less than P1000 and P1000 and over); (d)
education of mother (primary and below, elementary, high school, and college and over). A three-level
categorization of education is commonly adopted in the international literature: achieving literacy,
completing elementary education and completing secondary education and proceeding to higher education.
However, given the high level of literacy in the Philippines (above 80 per cent), a considerable proportion
(about 50 per cent) with elementary education, a sizable proportion proceeding to high school and no
further, and a quite small proportion continuing to higher education, a four-level categorization might
provide clearer child mortality differentials by education in the Philippines. The primary and below
category corresponds very well to illiteracy and bare literacy. Because a large proportion of those who
succeeded in attaining high school education could not proceed to college, perhaps because of financial
problems, it would be worthwhile to discover how high school education compares with elementary
education in explaining child mortality.
Health-related factors: (a) source of drinking water supply (unsa/edake or river, stream, spring,
rainwater, open well, pump shallow well; and safe: artesian deep well, pipe water); (b) presence of toilet
(none, outside the house, inside the house); (c) housing quality (inadequate: walls made of scrap materials,
nipa, other thatch, sawalit bamboo, rough-hewn timber and/or poorly-fitted planks and floors being earth or
constructed of bamboo, cement, and wood; and adequate: walls made of painted and /or well-fitted wood
or hollow blocks, cement or other expensive materials and floors of wood, linoleum or tiles); and (d)
household composition (extended and nuclear).
4
The socioeconomic and health related-environmental variables were characteristics at the time of the
survey. They might or might not refer to the characteristics while the child was exposed to the risk of
death. This problem might be more severe for the health related-environmental variables. For example, the
source of drinking water supply, housing quality and toilet facility might have been recently upgraded such
that some of the children classified as having better facilities might have been exposed to poorer facilities,
thus resulting in a bias towards mortality risk higher than was the case. These limitations must be kept in
mind when considering the results. Since they were the only available variables to consider, it was
assumed, as what other studies have been forced to do (e.g. Trussell and Hammerslough, 1983; Casterline
et al., 1989), that these covariates did not vary with time. In fact, there has been no remarkable and
persistent rapid economic change in the Philippines over time (Hill, 1986, 1988; Herrin, 1988); hence, it
may be safe to assume that most of these variables might have been stable over time for most of the
population in question.
Cabigon (1990) showed that infant and child mortality rates per 1000 births have levelled off since 1960.
To allow for the higher rates before 1960 in these indicators and to include all births in the analysis, a
variable for time period of birth was created by dichotomizing the births by occurrence into those before
1960, and 1960 and later.
The outcome variables examined here were neonatal, post-neonatal, infant and child mortality. Several
studies revealed that factors affecting neonatal and post-neonatal mortality vary greatly, the former being
more often biological, medical, and congenital and the latter more often social and environmental
(Antonovsky and Bernstein 1977; Cleland and Sathar 1984; Cramer 1987; Gortmaker 1979; Hobcraft et al.
1983, 1984; Palloni and Tienda 1986; Pharoah 1976; Pharoah and Morris 1979; Shapiro, Schlesinger and
Nesbitt 1968; Trussell and Hammerslough 1983). So it is pertinent to look separately at the two
components of infant mortality, neonatal and post-neonatal, in this study. However, there may be some
factors which do not appear important in either the neonatal or post-neonatal disaggregations, because of
the small number of cases, but may be influential when infant mortality as a whole is considered, because
of a sufficient number of cases. Estimates of coefficients based on small number of deaths tend to be
erratic, indicating the existence of random variations. Hence, overall infant mortality was also examined.
The model
The log-linear rate or hazard model fitted by using the program GLIM 3.77 (Payne, 1985) was the
statistical tool used to fulfil the above objectives. Basically, with a log-linear rate model, the ratio or rate
(rty) of the total number of deaths (£>-) of children with a particular set of characteristics, or in the ith row
and jth column of a contingency table to the total amount of exposure (N-) of children with those
characteristics is considered (Hobcraft et al., 1984:209-210; T. W. Pullum, personal communication. May
1, 1989). This model is based on the assumption that there is homogeneity among children with similar
characteristics and that the number of deaths follows a Poisson distribution with expectation calculated as:
E^Dip-riJNij-
(1)
It was also assumed that the effects of the explanatory factors on the rate are multiplicative and that each
rate is composed of a product of a constant term (c), a row term (a-), and a column term (d-). That is.
(2)
One of the complexities of all log-linear models is over-parameterization or too many parameters to
estimate. In an over-parameterized model, individual parameters cannot be estimated unless some
constraints are added to the model. In GLIM, the constraint imposed is that all parameters, with one or
more index equal to one, are set to zero. This restriction imposed in GLIM is termed a regression-like
constraint by Long (1984) or a dummy coding effect by Alba (1987). These parameters set to zero refer to
the omitted or reference categories. Equivalently, to obtain estimates of parameters and to define the
parameters uniquely, the restriction imposed is to set:
a^b^l so that r^-c and the model can be re-expressed as
(3)
In words, the cell in the table, corresponding to the row and column parameters set equal to unity, is
called the baseline or reference cell. The parameters ai and bj are interpreted by using the epidemiological
concept of relative risk, i.e., a-as the relative risk associated with an individual being in the zth row relative
to an individual being in the first row or baseline category and bj to be interpreted analogously.
5
Transforming the multiplicative rate model into a linear equation by taking natural logarithms of all the
terms eases its estimation. The equation is linear in its logarithm, hence called log-linear. Thus,
logE^D ^=10gNy+M+A -+By
(4)
where log/Vy=offset (a quantitative variate whose regression coefficient is known to be 1)
M=constant term called the overall mean
A^ysrow and column effects, respectively, due to the qualitative factors in question.
Exponentiating the log-linear parameters restores the multiplicative form of the rate model. Since the
model is log-linear rate, the exponentiated parameter estimates are relative rates. Any value less or more
than unity means, respectively, lower and higher relative risk of dying at the age interval in question of the
group under consideration than that of the baseline group. The exponential of the constant or overall mean
is the fitted rate for the reference group, which is the reference rate. Multiplying this rate by the relative
rates of the considered categories yields probabilities of dying or fitted rates for such groups (Hobcraft et
al., 1984; T. W. Pullum, personal communication, July 28,1989).
Testing and model selection
To deal with the numerous available variables listed earlier and their high interrelationships (e.g. current
residence, presence of electricity, household income and education) required two analytical steps. First was
to identify those that remained significant after controlling for a number of other variables. This step
produces main effects models (models with the minimum number of predictors). Log-linear analysis
emphasizes goodness-of-fit of a model and often, the main effects model is not the optimal or best fitting
model that provides the best explanation of the observed relationships between the explanatory factors and
child mortality. Hence, the second step was to test for interaction effects.
In essence, determining the total and specific effects of covariates requires a systematic examination of
interactions. When an independent variable does not interact with other independent variables, the specific
effect for that variable equals its total effect and the magnitude of the effect does not depend on the levels
of the other variables. The exponentiated estimates or relative risks are then interpreted in a
straightforward manner. When the variable in question interacts with other independent variables, the total
and specific effects of that variable depend on the level of the variable or variables with which it interacts.
If it interacts with just one independent variable, then its total effect depends on the level of that variable
and its specific effect is the ratio of total effects for the given levels of that variable. If it interacts with two
other independent variables, then its total effect depends on both the levels of these two variables and its
specific effect is the ratio of total effects for both levels of these two variables; thus the interpretation
becomes more complicated. The higher the order of interaction effects, the greater the complications in
interpreting the results of the log-linear analysis.
There are two extreme approaches to derive main effects or optimal models. One is the forward approach
which starts with the simplest model and builds up the model by adding variables or terms one at a time.
At each stage, the variable or term that makes the greatest improvement, according to some criterion, is
brought in to produce the next model. The process is then repeated. In this approach, in the early stages,
the model being tried is incomplete: for example, important regressor variables will not yet have been
included in the model. This is its major drawback, the possibility of not being able to examine fully the
importance of all potential regressors and hence the likelihood of some variables, though important, not
being included, especially when the starting point is an arbitrary choice among several radically different
possibilities.
The other is the backward approach which starts with all variables and terms (saturated model) in the
model and systematically eliminates variables or terms according to some criteria. Having to start with all
potential regressor variables or terms, the backward approach is not faced with the difficulty associated
with the arbitrary choice among interrelated possibilities because it starts with all possibilities and
eliminates irrelevant variables or terms. In this manner, no potential regressor is likely to be missed,
especially if the saturated model is correctly specified. However, the important problem of this approach is
that with so many possibilities to start with, it tends to involve only automatic selection procedures and
because of too many parameters to estimate, in the early stages big models may not be very well fitted,
making it difficult to distinguish between important and less important variables or terms. Nonetheless,
despite its major limitation, the backward approach , as Cox and Snell (1981, p. 22) wrote,
is the safer one and should normally be used when it is not too ponderous and especially when there is a
major interest in and uncertainty over the primary formulation of the problem. The forward approach is more
appropriate for the secondary aspects of the problem, e. g. over the structure error.
6
A synthesis of these two approaches is the stepwise approach which seeks to fulfil both the elimination
and addition of variables or terms according to some criteria. Nonetheless, in log-linear models, where
there are-too many parameters to estimate, with many explanatory variables to choose from and a very
large number of observations, stepwise and backward model selections require enormous cost, time and
computer workspace. Most existing software packages cannot handle the problem. For example, GLIM
does not allow stepwise derivation of optimal models. With 42,471 births to deal with, a consideration of
all these points led to the adoption of the forward selection in the derivation of the main effects models. I
used both forward and shortcut backward6 approaches to identify significant interaction terms to arrive at
the optimal models.7 I pegged up the choice of the optimal model to third-order interaction parameters to
avoid very complicated interpretations of results. I based the search for the optimal models on condensed
files created by the GLIMTAB .FOR FORTRAN program written by T. W. Pullum (personal
communication. May 1, 1989), as the 42,471 births taken individually could not be handled. However,
calculating the deviances and degrees of freedom of the more complex models and the optimal models
based on the individual files is straightforward once the baseline or null model deviance is calculated from
the individual files. The DEVBASE.FOR FORTRAN program of T. W. Pullum (personal communication.
May 9, 1989) yields both this deviance and its corresponding degrees of freedom al±ough the latter is
simply the total number of cases minus one. The appendix contains a detailed treatment of these two
programs and the procedure of calculating the deviances and degrees of freedom of the more complex and
optimal models based on the individual files.
Use of weights
As stated earlier, stratified sampling was employed in the probability of selecting the sample with
barangay and household as the primary and ultimate sampling units, respectively. Weights were then
assigned to each record in the survey to reflect the sampling proportions. Issues on the use of weights in the
analysis of survey data were also noted above. The suggested solutions by Lee et al. (1986) and Clogg and
Eliason (1987) and the currently available software packages that address such issues require extremely
complicated approaches. So I chose the conventional strategy and as a compromise, as proposed by
T. W. Pullum (personal communication. May 1, 1989), I used the weights for estimating parameters but not
for testing and selecting the main effects and optimal models.
There is no question that fit statistics based on unweighted data disregarding the sampling design are less
biased relative to the weighted data (conventional approach). But the biases that are raised may not be
serious in some cases. One important point is that such biases are dependent on the range of the weights
used. If a considerable proportion is weighted by a value close to unity, the biases may not produce serious
problems. In the 1983 NDS, the weights range from 0.1734 to 2.0731 with around 33 per cent of the cases
weighted by values close to one. This may not produce serious biases. Hence, the adopted approach of
using the unweighted data in the testing and model selection in this analysis, though biased, may still be
robust
Results
Univariate analysis
Demographic factors. Table 1 presents the demographic, socioeconomic, and health-related differences
in neonatal, post-neonatal, infant, and child mortality. Both demographic variables showed some
relationships with each of the dependent variables: neonatal, post-neonatal, infant and child mortality.
Birth order and mortality at ages beyond the first month of life showed a consistently positive association.
During the neonatal period, first births were more likely to die than higher order births.
Table 1 about here
Children bom to mothers aged less than 20 years were more likely to die in their first five years of life
than those bom to older mothers. Those bom to mothers at high-risk ages, 35 years and over, did not differ
in mortality during the first month and second to fifth years of their lives from those bom to mothers at the
middle ages, where associated risks, such as congenital anomalies, deterioration of uterine efficiency and
difficult and prolonged labour, are expected to be generally low. However, it is during the post-neonatal
and overall infancy periods that those bom to mothers at die oldest ages of maternal delivery were more
likely to die than those bom to mothers at the middle age group.
Socioeconomic factors. All the socioeconomic variables manifested a clear negative relationship with
each of the child mortality indicators in question, with education of mother emerging as the most important,
followed by income, then presence of electricity, and current residence the least important. The relative
importance of mother’s education has been consistently established in related studies of the Philippines as
well as other developing countries. The value of the four-level categorization of education was seen with
child mortality as the mortality indicator, where the child mortality rate associated with the high-school
7
category was twice as low as that relating to the category of elementary education. The clear univariate
mortality differential by income might indicate that a neglect of this variable in the causal chain leading to
child survival might exaggerate the effect of other socioeconomic, factors, especially maternal education,
which has been identified as a major socioeconomic determinant. The magnitude of the child mortality
differential by current residence and presence of electricity in the household, other proxies for the
socioeconomic status of the household, was not as large as that indicated by education and income.
Health related-environmental factors. Only the presence or absence of a toilet and housing quality had a
clear and consistent relationship with child mortality at any age. That is, those belonging to households
with unsanitary conditions (no toilet and inadequate housing facilities) were more likely to experience child
mortality at any age than those living in a sanitary environment. Source of drinking water was slightly
associated with mortality between the first and fifth years of life. Household composition was not related
with any of the dependent variables.
Multivariate analysis
The observed bivariate associations warranted further investigation because most of the variables were
interrelated. The next task then was to consider which of these variables constituted the model that best
explained the outcome variables in question. This led to the search of the optimal main effects and final
optimal models that best described the relationships between these variables and neonatal, post-neonatal,
infant and child mortality.
Table 2 summarizes the null, main effects and optimal models for each dependent variable and their
corresponding deviances and degrees of freedom resulting from the modelling exercise. Because the
forward model selection was adopted to arrive at the main effects models, the variables were arranged
according to importance, with the first as the most important and the last, the least important, in each of the
main effects models. Nevertheless, it is important to note that the order of importance becomes
meaningless when some variables significantly interact with each other. It is only when there are no
significant higher-order interactions that ranking the variables by importance holds true, as evident with
post-neonatal mortality8 Hence with post-neonatal mortality, the interpretation of the effects of each of the
predictors was straightforward since the effects on the outcome of post-neonatal mortality were direct.
This means the effects did not depend on the levels of the other variables. For the other outcome variables,
the optimal model was the main interest, in which the effects of interacting variables differing for the
various levels of the other variables they interacted with were interpreted accordingly. The interpretation
also considers that an interaction mainly cancels out large main effects.
Table 2 about here
The order of interpreting the effects of a specific variable accorded with its order in the optimal model,
when it did not interact significantly with another variable. Where there were interactions, the predictors
with no significant interactions were interpreted first. The next were the two-way interaction terms and the
last, the three-way interactions. The following is a discussion of each of the optimal models treated
separately under each of the components of child mortality.
Neonatal mortality. From Table 2, out of the 11 variables tested, the forward model selection method
identified the length of the preceding birth interval (P), education of mother (ED), birth order (BO) and
maternal age at childbirth (M) as the best predictors of neonatal mortality. Significant higher-order
interactions among P, ED and M emerged. Table 3 presents the parameter estimates and their
exponentiated values, which were termed relative rates with the exponentiated value of the overall mean as
the reference rate, of the main effects and optimal models. Since birth order did not interact with the other
predictors in the optimal model, its specific effect was its total effect on neonatal mortality. Its effect (Part
A) was then dealt with first before the interacting effects (Part B) of the other important variables. To
simplify the analysis of the interaction effects, the derived log-linear coefficients were occasionally
interpreted. If the coefficient is negative, then the group has lower mortality than the reference group.
Table 3 about here
Thus, according to the optimal model, the chance of dying during the first month of life for a second or
higher order birth was about 33-36 per cent lower than that for first births. Clearly, first births were most
likely to experience the highest neonatal mortality in net terms.
The effect of the other health-risk factor, maternal age at childbirth, on neonatal mortality varied
according to the length of preceding birth interval and mother’s education. For all interacting variables, the
reference was arbitrarily those births belonging to the shortest preceding birth interval, youngest maternal
age at childbirth and lowest education of mothers. These births are most likely to be exposed to the highest
8
relative risk of neonatal mortality. The interaction among maternal age at childbirth, mother’s education
and preceding birth interval shed further light into the observed main effects of the first two variables of
interest. Effects were notably and generally large, although most deviations are not statistically
significantly different from zero, at higher levels of mother’s education at the shortest preceding birth
interval and at any given level of mother’s education at preceding birth intervals of 18 months and over.
Effects were notably small at levels below college education of mothers at the shortest preceding birth
interval.
Nevertheless, the inverted-J-shaped pattern of effect of maternal age at childbirth observed in the main
effects model was evident at most levels of preceding birth interval and mother’s education with the gap
between the less than 20 and 20-34 years of maternal age at childbirth, ranging from 25 to 77 per cent For
example taking those statistically significant values, for babies whose mothers had primary or no education
and with 31 months or over of preceding birth interval, the expected neonatal mortality rate for those
delivered at ages 20-34 years of their mothers was 36 per thousand (exp(-1.33)*0.137), which is 25 per cent
lower than the corresponding rate of 48 per thousand (exp(-1.05)*0.137) for those delivered at very young
maternal ages. Those deviating from this pattern fell under the college or over educated mothers-all levels
of preceding birth interval, the lowest educated mothers-shortest preceding birth interval and elementary
educated mothers-18-30 months preceding birth interval groups. The very large standard errors of most of
these deviant estimates indicate their instability as they were based on very few cases; therefore patterns
manifested were inconclusive.
Moreover, the non-monotonic decrease of neonatal mortality with mother’s education observed in the
main effects model became more interpretable with the optimal model. Those that departed most from the
expected inverse relationship between neonatal mortality and mother’s education related again to most of
the college or over educated mothers group where cases were very few. Strikingly however, at any given
level of preceding birth interval and maternal age at childbirth, except the shortest preceding birth intervaloldest maternal age at childbirth, those babies of elementary-educated mothers experienced neonatal
mortality much lower than babies of mothers with no or primary education. In fact, for the extreme
preceding birth interval-youngest maternal age at childbirth and 18-30 months preceding birth
interval-20-34 years maternal age at childbirth categories, the higher the education of the mother up to high
school, the lower the expected neonatal rate.
The predictive power of the optimal model for neonatal mortality was clearly illustrated by contrasting
the predicted neonatal mortality rates of two extreme groups. For example, the predicted neonatal mortality
rate for the baseline group, first births with less than 18 months of preceding birth interval, delivered at the
youngest ages by the lowest educated mothers was 137 per thousand, which is 82 per cent higher than than
the predicted rate of 24 per thousand (0.137*0.670*0.264) for second or third births with 31 months and
over of preceding birth interval delivered at ages 20-34 years by the same lowest educated mothers.
Post-neonatal mortality. For the post-neonatal mortality, all the health-risk or demographic variables preceding birth interval (P), birth order (BO) and age of mother at birth of the child (M) - were highly
significant covariates, with P and BO ranking first and second (Table 2). Among the socioeconomic
variables, average household monthly income (IN) was more important than mother’s education in
affecting post-neonatal mortality. The inclusion of household income and mother’s education in the
modelling process eliminated the importance of both current residence and presence of electricity in the
household. Time period of birth (BI) was important in delineating the post-neonatal experience between
infants bom in the distant past and those bom in the recent past, with the latter showing lower post-neonatal
mortality. Among the health-related variables, only housing quality (H) and toilet facility (T) showed
significance.
Turning now to the impact of the significant predictors of interest, with P and BI as controls (Table 4), all
other things being equal, the second or third order births had a lower rate (about 25 per cent) of postneonatal mortality than first births. Although the effect estimate for fourth or higher order births is not
statistically significantly different from zero, it implied a higher relative risk for this group than first births.
In effect, first births and fourth and higher order births were more likely to be of greater risk to postneonatal mortality than second or third births after controlling for all the other variables.
Table 4 about here
The higher the average household monthly income, the lower the chance of dying between the first and
twelfth months of life. The expected relative rate for the more economically advantaged group was about
19 per cent lower than for those in the less economically advantaged group.
The net effect of education indicated that those bom to mothers with college or higher education would
experience post-neonatal mortality about 47 per cent lower than those bom to mothers with no or primary
i
9
education. Elementary and high school education reduced the risk of post-neonatal deaths by a little above
30 per cent
Those delivered when their mothers were in their twenties and early thirties were less likely to die (about
30 per cent lower) than those delivered by mothers at ages below 20 years. Although the estimated effect
for the oldest age at childbirth group is not statistically significantly different from zero, it indicated that the
relative risk was lower for this maternal age group than for the youngest age of delivery group.
Babies bom to mothers with adequate housing facilities were less likely by 15 per cent to die during the
post-neonatal period than those bom to mothers with inadequate housing facilities. Presence of toilet, at
least outside, reduced the risk of post-neonatal deaths by 21 per cent Again, the parameter estimate
referring to the category ‘inside* is not significantly different from zero. However, the direction of its
effect was consistent with expectations.
Contrasting the estimated relative risks of the most advantaged and disadvantaged groups illustrated the
predictive power of the best-fitting model for post-neonatal mortality. That is, the expected post-neonatal
mortality rate for the second or third child bom to a mother aged 20-34 years in a household with an
average monthly income of one thousand or more pesos, with college or higher education, with adequate
housing quality and with at least a toilet facility is about 28 per thousand
(0.188*0.748*0.698*0.811*0.527*0.852*0.787), which is a little more than six times lower than that of a
child in the baseline group (188 per thousand), namely a first birth to a mother aged below 20 years in a
household with an average monthly income of less than a thousand pesos, with no or primary education,
with inadequate housing quality and with no toilet facility.
Infant mortality. With overall infant mortality (Table 2), all the significant predictors of post-neonatal
mortality except housing quality (H) maintained their importance. Source of drinking water (W) replaced
H and the optimal model involved second-order interactions between P and M; P and W and third-order
interactions among ED, T and W.
Table 5 displays the effects of these significant predictors of infant mortality. The predictors that did not
interact with other variables were birth order and household income (Part A). As observed with neonatal
and post-neonatal mortality, the relative risk to infant mortality of second and third order was 33 per cent
lower than that of first births. Fourth or higher order births had a slightly lower relative risk than first births.
As in post-neonatal mortality, the relative rate of the higher income births was 19 per cent lower than that
of lower income births.
Table 5 about here
Maternal age at childbirth interacted with length of preceding birth interval. The shortest and longest
preceding birth interval groups showed the smallest and largest effects, respectively, a pattern similarly
observed with neonatal mortality (Part B). At any given level of preceding birth interval, the effects of
maternal age at childbirth followed the same inverted-J-shaped pattern clearly evident with post-neonatal
mortality and less clearly with neonatal mortality.
Among the three interacting variables (Part C), the interpretation took education first. Interaction effects
of mother’s education with toilet facility and drinking water source cancelled out the observed large main
effects of mother’s education. Educational differentials were greater among babies having unsafe source of
drinking water than among babies having safe source of drinking water at any given level of toilet facility.
Taking for example the no toilet facility-unsafe drinking water supply group yielded a difference of 43 per
cent in the fitted rates between births of the lowest educated mothers, 353 per thousand, and those of
elementary-educated mothers, 200 per thousand =0.353*exp(-0.57). On the other hand, the difference in
the expected relative rates of births in the no toilet facility-safe drinking water supply group between the
same two lowest educational levels of mothers was negligible. Also, the gap in relative rates between those
with no or primary-educated and high school educated mothers but having the same unsafe source of
drinking water were 54 and 41 per cent, respectively, for outside and inside categories of toilet facility. In
contrast, the corresponding differences in relative rates between the compared educational groups but
having the same safe source of drinking water were 16 and 10 per cent, respectively, for outside and inside
categories of toilet facility.
The net inverse relationship between mother’s education and infant mortality observed in the main effects
model was evident only at the outside toilet facility-unsafe water source and inside toilet facility-safe water
source levels. The remaining levels showed less clear patterns, although the general patterns of lower
relative risk of births of elementary-educated mothers compared to that of births of the lowest educated
mothers and of the lowest relative risks of births of mothers with the highest educational attainment
prevailed.
10
On the effects of toilet facility, the log-linear significant coefficients consistently increased from -0.57 to
-1.17 among those with an unsafe source of water and whose mothers had elementary education. The same
pattern, although not uniformly consistent, held true with the other levels of education and source of
drinking water, suggesting that among the three interacting variables, toilet facility had the sharpest impact
on infant mortality.
In the main effects model, the net effect of source of drinking water followed the unexpected direction of
causation. This was mainly because of the larger effects observed in the cells referring to the category
‘unsafe’ compared to the cells referring to the category ‘safe’ at the levels of: (a) no toilet-elementarycollege-educated; (b) outside and inside toilet-high school-educated; (c) inside toilet-elementary-educated;
and (d) outside toilet-highest educated mothers. For these groups, the infants with a safe source of drinking
water were likely to experience higher infant mortality than those with an unsafe source of drinking water.
This is a puzzling finding. Nonetheless, the remaining levels manifested the expected direction of lower
infant mortality, the safer the source of drinking water in net terms. For example, among those whose
mothers had no or primary education, irrespective of type of toilet facility, or whose mothers had high
school education with no toilets, those with safe source of drinking water had lower relative rates than
those with unsafe source of drinking water.
The unexpected direction of causation between drinking water supply and infant mortality in some of the
levels of toilet facility and mother’s education could be due to the dichotomous categorization of water
supply source. However, examining the infant mortality rate by education and source of drinking water,
with its most detailed categorization, in Table 6, revealed that among the elementary educated mothers, the
infant mortality rate of those obtaining their drinking water from pipes was higher than most, if not all, of
the other categories in question in both urban and rural areas. Regardless of current residence, a pump
shallow-well water supply, which was expected to be more unsafe than artesian deep well or piped water,
and rain water implied lower infant mortality than did the other sources of water. If misreporting of
sources of drinking water was the major reason for the unexpected pattern, the reverse pattern would have
been observed among the higher educated mothers. Nonetheless, the same lower infant mortality rate for
such sources relative to the others was observed.
Again, comparing the fitted rate of 19 per thousand (0.353*0.670*0.811 *0.239*0.419) for second or
third births belonging to household income of P1000 or more (monthly), delivered by high school educated
mothers at ages 20-34 years, with preceding birth interval of 31 months and over, with toilet outside the
house but with unsafe source of water, with the fitted rate of 307 per thousand (0.353*0.869) for fourth or
higher order births of mothers with low income delivered by mothers of no or primary education at ages
below 20 years, with preceding birth interval of less than 18 months, no toilet and unsafe source of drinking
water yields a difference of 94 per cent This demonstrates how powerful was the optimal model for infant
mortality.
Table 6 about here
Child mortality. Child mortality (Table 2) had the same eight significant covariates as post-neonatal
mortality. The only difference was the way the effects of some of these predictors operated. While the
optimal model for post-neonatal mortality was the main effects model, that for child mortality included
more second-order and third-order interactions.
Table 7 shows the resulting parameter estimates and relative rates for the main effects and optimal models
for child mortality. Interpreting the effects of the predictors with no interactions (Part A) yielded a clear
and consistent impact of maternal age at childbirth and housing quality on child mortality. The predicted
child mortality rates for those bom to mothers aged 20-34 and 35+ years were respectively 22 and 34 per
cent lower than for those bom to mothers below 20 years. The inverted-J-shaped pattern of relationship
between neonatal, post-neonatal and infant mortality and maternal age was not evident with child mortality.
The higher the maternal age, the lower the child mortality. Those living in adequate housing were 28 per
cent less likely to die during their childhood years than those in inadequate housing.
Table 7 about here
Part B presents the interaction effects of birth order and household income. In the optimal model, birth
order interacted both with preceding birth interval and household income. Given that preceding birth
interval was treated as a mere control, the interactions between birth order and household income were
more relevant. Thus, regardless of income level, the fourth order births experienced the highest relative risk
of child mortality. However, among low income-births the effect of birth order followed a J-shaped
pattern. The relative rate for low income fourth or higher order births was 112 per thousand (0.082*1.363)
while that for low income second or third births was 58 per thousand (0.082*.705) and that for first births
(reference group) was 82 per thousand. Interestingly, for the higher income group, the higher the birth
order, the higher the relative risk of child mortality. The relative rate for first births was 49 per thousand,
for second or third order births, ’ll per thousand and for fourth or higher order births, 100 per thousand.
11
Part C shows the coefficients from the three-way interactions. As had been done with neonatal and
overall infant mortality, the effects of these interacting variables were occasionally assessed using the
log-linear coefficients.
The effects of income depended on both the level of birth order and the levels of education and toilet
facility, so they were interpreted first on the level of birth order and then on both the levels of education
and toilet facility. The ratios of relative rates between household income category <P1000 and category
Pl000+ were 1.68 and 1.12 for first and fourth or higher order births, respectively, indicating higher
mortality for the poorer group. According to the three-way interaction terms in C, although some
coefficients are not significantly different from zero, for any given level of education and toilet facility, the
lower the risk of child deaths, the higher the average income of the household. This indicated a clear and
consistent negative association between household income and child mortality, net of the confounding
effects of all other important variables.
There was an inverse relationship, although not as clear and consistent as that observed with household
income, between mother’s education and child mortality. Among those with a toilet outside the house and
a lower income level, ±e higher the educational attainment, up to high school, the lower the child
mortality. In fact, if statistical significance is disregarded, for those with toilets inside the house, the higher
the household income and education of the mother, the lower the risk of child mortality. For those with
toilets outside the house, the same pattern existed for levels of education above primary. The only
exception to this pattern occurred to births to college educated mothers with no toilet and low household
income. The estimate, although not significantly different from zero, meant that those belonging to
mothers with the highest educational attainment, below average income level and no toilet were likely to
experience the highest mortality (log-linear coefficient=0.27). However, since there were only 98 births in
question and the estimate is not statistically significant, the pattern observed should be treated with caution.
The effects of toilet facility were more or less in the expected direction, although less clearly than
income. The general pattern observed was that having at least a toilet was associated with lower child
mortality. For example, taking the significant estimates, among the births to high school educated mothers,
the better the toilet facility the lower the child mortality as seen in the increasing log-linear coefficients
from -0.58 to -2.19.
The estimated child mortality rate for second or third order births delivered by college educated mothers
at ages 20-34 years with adequate housing quality and toilet outside but low income was 6 per thousand
(0.082*0.779*0.719*0.705*0.186). This value is 88 per cent lower than the rate of 49 per thousand
(0.82*0.657*1.363*0.664) for fourth or higher order births to mothers of no or primary education at ages
35 years or more with inadequate housing quality, toilet outside but low income. This again indicated how
fit the optimal model was to the child mortality data.
Discussion
The present log-linear rate analysis of covariates of child mortality has identified demographic,
socioeconomic and health-related factors affecting neonatal, post-neonatal, infant and child mortality in the
Philippines using the 1983 NDS. The importance and patterns of effects of these variables on each of the
components of child mortality varied to a certain extent. Hence, a discussion of the main findings and their
corresponding implications under each component is helpful.
(1) The ‘best’ model that predicted neonatal mortality included all the demographic or health-risk
factors
preceding birth interval, crudely measured, birth order and maternal age at childbirth----- and
mother’s education. All other things being equal, the pattern of net effects of birth order and maternal age
at childbirth was an inverted J-shape, with first births or any birth to mothers at their youngest ages of
reproductivity experiencing the highest risk, followed by fourth or higher order births or any birth to
mothers at their oldest ages of reproductivity. Differences by maternal age at childbirth were great with
preceding birth intervals of less than 18 months, regardless of mother’s education and with preceding bmth
intervals of less than 31 months and no or primary education of mothers. Also, educational differentials
were marked at the youngest maternal ages at childbirth and shortest preceding birth interval. These
patterns identify the groups most likely to experience higher than average neonatal mortality. They are first
births and any birth at very young ages of mothers with no or primary education and with very short
preceding birth intervals. They are targets requiring top priority in the implementation of health programs.
The prominence of all three health-risk factors
precedmg birth interval, birth order and maternal age
at childbirth
reinforces findings of several studies that biological or medical rather than environmental
factors are associated with neonatal mortality. However, the emergence of education as an important
variable in the ‘best’ model may imply that the dominance of the biological or medical factors in affecting
neonatal mortality applies mainly to high-income countries. In low-income countries, like the Philippines,
12
there tend to be socioeconomic differentials in neonatal mortality, as evidenced by the significant impact of
education of the mother. Nevertheless,the interaction effects of mother’s education with preceding birth
interval and maternal age at childbirth indicates that formal schooling as such is not the key determinant of
neonatal mortality. Although births to highly educated mothers tended to show the lowest risk of neonatal
mortality, variations by mother’s education at longer preceding birth intervals and older maternal ages at
delivery were small and non-uniform. These patterns imply that mother’s education may be significant in
its own right, or may be a reflection of differentials owing to education in nutrition of the mother, care of
the mother during pregnancy or in conditions of maternal delivery.
(2) The best-fitting model for post-neonatal mortality is consistent with the theory that socioeconomic and
health related-environmental factors become more prominent after the first few months of life.
Demographic or health-risk factors, length of preceding birth interval and birth order, remained the most
important predictors; the socioeconomic factors, average household income and mother’s education, ranked
next; and health-related factors, housing quality and presence of toilet, came last. It may be argued that this
order of importance of the predictors is biased through the forced assumption that the socioeconomic and
health-related covariates did not vary with time. Nonetheless, confining the analysis to births occurring
five years before the survey did not change the above ranking (Cabigon 1990).
That the direct effect of average household income was greater than that of mother’s education implies
that formal schooling, which has been shown in previous Philippine studies as the strongest socioeconomic
determinant, may have reflected income characteristics and that maternal education may not be an adequate
proxy of Philippine socioeconomic status. In addition, increasing the purchasing power of the populace
may need attention equal to or even more than improving their educational levels.
The patterns of effects of these ‘best’ predictors of post-neonatal mortality show that post-neonatal
mortality was high for babies: (a) in the first and fourth or higher order; (b) delivered at very young and
old ages of their mothers; (c) belonging to low income households; (d) with low-educated mothers; (e)
living in an inadequate housing environment; and (f) in households lacking toilet facilities. These are the
critical groups to be given top priority in health planning and implementation if the goal is to reduce
Philippine post-neonatal mortality.
(3) The most parsimonious model for infant mortality identified the same significant demographic and
socioeconomic covariates as in the post-neonatal mortality model. However, while the net effects of these
predictors on post-neonatal mortality were uniform across all subgroups of the population in question, they
were not so with overall infant mortality. Only birth order and average household income maintained their
direct effects on overall infant mortality. The rest showed varying effects on various levels of the other
predictors they had interactions. Also, both models differed in the identification of important health-related
predictors. Housing quality was important for post-neonatal mortality but not for the overall infant
mortality. It was replaced by source of drinking water in the infant mortality model. It must be noted
though that source of drinking water was only influential with overall infant mortality, a finding consistent
when provinces were units of analysis (Cabigon 1990) and with the finding of Martin et al. (1983) of its
insignificance on child mortality.
The observed dissimilar patterns of covariates of post-neonatal and overall infant mortality suggest the
necessity of examining infant mortality, in both its disaggregated and overall forms. Analysing its
components is important, as many factors are associated with different risks at different ages of the infant.
Examining overall infant mortality may show some patterns obscured when its components are
investigated, because of smaller sample size, which had been the case with the present data.
Turning to the patterns of effects of these predictors of infant mortality reveals the same patterns of
effects of birth order on neonatal mortality and of household income on post-neonatal mortality. That is,
first and fourth or higher order births and low-income births were likely to experience very high infant
mortality. As with neonatal mortality, maternal age at childbirth interacted with preceding birth interval.
For the shortest and longest preceding birth interval, those delivered at the youngest maternal age were
most likely to have the highest risk, followed by those at the oldest maternal age at delivery. The reverse
held true for births with 18-30 months preceding birth intervals.
The effects of education depended on the level of source of drinking water. Differences by mother’s
education were greater with unsafe than safe source of drinking water. Among those with unsafe sources
of drinking water, the expected pattern held, although not very consistently: that is the higher the mother s
education, the lower the infant mortality, at any given level of toilet facility. Among those with safe source
of drinking water, the net inverse relationship was less clear although the general pattern was in the
expected direction.
13
Net effects of toilet facility were, however, in the expected direction of negative association with infant
mortality at any given level of education and source of drinking water. In fact, among the three interacting
variables, mother’s education, toilet facility and source of drinking water, toilet facility showed the sharpest
association with infant mortality within the explanatory framework used.
It is the effect of source of drinking water that displayed expected and unexpected patterns at different
levels of mother’s education and toilet facility. The unexpected pattern persisted even with the most
detailed categorization of source of drinking water.
Several explanations of the unexpected pattern are advanced. First is misreporting of the type of drinking
water source. It may be possible that unsafe sources were reported as safe by some respondents in the
survey. However, the detailed categorization of drinking water supply by infant mortality level did not
show a systematic bias towards reporting sources of drinking water as safe even if they were not.
Nonetheless, further investigation of this aspect is important before reaching definitive conclusions.
Second is the role played by behavioural practices enhanced by non-formal or formal education as clearly
indicated by the marked educational differentials by unsafe source of drinking water. Perhaps, knowing
their sources of drinking water are unsafe, most mothers with at least an elementary education may have
been boiling the water before consuming it, so it may be the behavioural practices rather than the source of
drinking water per se that were measured. This may be a more reasonable explanation than misreporting.
Third is the manner of transport from the source to the house and the means of storing the water. While
piped water and artesian wells were reported as sources of drinking water, these sources are likely to be
public sources for the majority of the population in question. Therefore, the container, the mode of
transport from the public source to the house, and the way of storing the water are important factors to be
considered for this segment of the population. In fact, the present study showed that as expected, the higher
educated mothers, who were more likely to afford tap water inside their houses, experienced fewer infant
deaths than those with other sources of drinking water, irrespective of current residence.
The last explanation is the possibility of contamination of water from rusting pipes or pipes contaminated
by floods. This particularly may be more serious in areas frequently flooded, for example the MetroManila area.
While the behaviour of mothers or the suggested contaminating factors may have played major roles, as
these are mere speculations, further research verification is needed. Moreover, the same points raised with
post-neonatal mortality, regarding the observed net effects of demographic factors and income, apply with
infant mortality. The persistent independent effect of income suggests that it is a significant determinant in
its own right. The interaction of education with the health-related factors indicates that apart from formal
schooling there are other attributes, like behavioural practices, which may be equally important
determinants.
(4) The optimal model for child mortality had exactly the same covariates as that for post-neonatal
mortality. However, as with infant mortality, the effects of most predictors varied by different levels of the
other predictors with which they interacted. Maternal age at childbirth and housing quality did not interact
with the other predictors in the model. The higher the maternal age at delivery and the better the quality of
housing, the lower the child mortality. The underlying biological mechanism associated with maternal age
during infancy is no longer a relevant issue at later childhood ages. Perhaps, after the first year of life, the
mechanism associated with maternal age may be more a reflection of how the child is cared for. It is likely
that older women are more experienced than younger women in caring for their children to prevent child
mortality. The finding on the significant role of housing characteristics in affecting child mortality
augments the few published works, as reviewed earlier. One possible explanation of the importance of
housing quality is that the mortality risk of already poor children is exacerbated by the poor quality of their
housing. Possibly, a poor chilcU living in inadequate housing conditions and sick with an infectious
disease, such as pneumonia and influenza, is likely to have a higher risk of mortality than a child, sick with
the same disease, but living in better housing.
On two interacting variables, birth order interacted with both preceding birth interval and household
income. Regardless of income level, the fourth or higher order births manifested the highest risk;
nevertheless, while among the low income group, first births ranked second, among the high income group,
second or third order births ranked next.
With the three interacting variables, household income, mother’s education and toilet facility, on child
mortality, income effects were the sharpest, followed by the toilet facility effects and last, the education
effects within the analytical framework used. The effects of income depended on the levels of birth order.
14
maternal education and toilet facility. For the first and fourth and higher order births, the higher the child
mortality, ±e lower the household income. The reverse held with second or third order births. For any
given level of education and toilet facility, the general pattern persisted of lower child mortality risks with
higher income. The effects of toilet facility more or less reflected the general pattern that having at least a
toilet was associated with lower child mortality, regardless of income and education levels. Disregarding
the only deviant cell based on very few cases, the effects of mother’s education also followed an inverse
pattern of association with child mortality. These illustrate the great role of household income, mother’s
education and health-related factors in the causal chain leading to child mortality. Having a college
education may be not enough, unless the attained education is a means of generating income. The issue of
producing college-educated mothers but not providing corresponding job opportunities arises. The impact
of education, operating on the levels of toilet facility and household income and showing less clear
associations than income, suggests the possibility of exaggeration in the measured effects of maternal
education on child mortality in previous Philippine studies.
On the whole, this paper clearly identified the groups at higher risk of each of the components of child
mortality. The outcome of targeting these groups in the implementation of health programs is undoubtedly
a better survival of Filipino children. The paper demonstrated the varying patterns of effects of
demographic factors (birth order and maternal age at childbirth), socioeconomic factors (household income
and mother’s education) and health-related factors (housing quality, toilet facility and source of drinking
water) on neonatal, post-neonatal, overall infant and child mortality. However, there still remain several
issues needing further inquiry such as those already raised in the discussion. One additional issue worth
investigating is whether the inclusion of proximate variables in the analytical framework changes the
patterns of effects of these emerging important but non-proximate variables.
15
Notes
^he authors cautioned that the toilet facility variable, being measured only at the time of the survey, may
not be a good proxy for the type of sewage disposal present when the children were exposed to risk.
2Trussell and Hammerslough (1983, p.14) used sample weights assigned per case and treated the
weighted sample as a simple random sample.
3The barangay is the smallest political unit in the Philippines.
4An evaluation of the 1983 NDS pregnancy history revealed that the problems common in many surveys systematic event misplacement toward the survey, and omission of events, especially by earlier cohorts and
in earlier periods - were not noticeable in most of the cohorts of women in the 1983 survey. As in previous
surveys, the errors observed might not have caused serious distortions in the maternity history. Therefore,
the 1983 NDS provided data suitable for assessing the size and relative influence of each of the
demographic, socioeconomic and health related-environmental variables on neonatal, post-neonatal, infant,
and child mortality (Cabigon, 1990).
5I recognize one main flaw of considering the preceding birth interval as a control. The preceding birth
interval was loosely defined for it included first births. Theoretically, the effect of birth intervals should
only be studied with birth orders two and above. It is only interbirth intervals that are relevant, where the
three mechanisms - foetal growth, milk diminution and resource competition - operate. First births
obviously cannot properly be assigned an interval since a previous birth. However, this study was not
focused on sorting out the processes producing the effects of the preceding birth interval. The
methodological problems inherent in the analysis of child-spacing effects on child mortality noted above,
calls for a separate investigation and I am currently working on it under a Rockefeller Foundation
postdoctoral fellowship. As around 80 per cent of the births analysed here belonged to birth orders two and
over and as first births are subject to somewhat higher than average mortality risks, it was considered more
illuminating to concentrate on measuring the influence of the other demographic, socioeconomic and health
related- environmental variables on neonatal, post-neonatal, infant and child mortality, taking into account
the length of the preceding birth interval. The strategy followed then was to assign the first births the
longest category interval to ensure that they were not characterized by a short preceding interval (and the
accompanying stresses).
6As suggested by T. W. Pullum (personal communication, June 16, 1989), the easiest, shortcut and
backward elimination from a three-way interaction can be obtained by comparing the deviances of the
three-way interaction model and its implied two-way interaction model. For example, given variables A, B
and C, the two nested models to be compared to identify third order interaction of A, B and C are: A*B*C
(which includes all the implied lower order effects or interactions) and A*B+A*C+B*C (the two-way
interactions and all the implied main effects).
7The deviance is a measure of how closely the model fits the data and is distributed asymptotically as
chi-square under the null hypothesis that the log-linear rate model and the underlying assumptions are
correct. Testing the significance of specific variables or with the addition of further terms was based on the
differences in deviances between the baseline model and a more complex model (nested models) compared
with the tabulated chi-square, with degrees of freedom (df) equal to the difference between the dTs of the
nested models under comparison. The df indicates how much information is available for estimating the
‘background noise’ with the residual variance (Swan 1985, p.18).
8A three-way interaction among household income, mother’s education and maternal age at childbirth is
significant at 0.05 level but adding this interaction term to the main effects model did not improve the fit of
the main effects model. The difference in deviance was 22.8 at 17 degrees of freedom, which is not
statistically significant. Moreover, the estimated interaction effects were practically unimportant
quantitatively for they slightly changed the main effects of each of the variables in question (results not
shown).
16
Acknowledgments
An earlier version of this paper is found in Chapters 8 and 9 of my Ph.D thesis submitted to the
Australian National University, Canberra. I accomplished this revised version while I was a postdoctoral
fellow of the Rockefeller Foundation at the Division of Demography and Sociology of the Research School
of Social Sciences, The Australian National University. Debts of gratitude are owed to several persons.
Substantial improvement of the earlier version was due to the guidance of my supervisors. Prof. Gavin
W. Jones and Dr. Christabel M. Young, my advisers. Dr. Alan Gray, Dr. Ian Diamond and Dr. Terence
H. Hull and a close associate. Dr. Thomas W. Pullum. Dr. Thomas W. Pullum patiently taught and assisted
me via mail communications. He wrote FORTRAN programs to create condensed files and to estimate
baseline deviances from individual files as the individual files are too large for GLIM to handle, especially
if more than 10 factors are considered. Dr. Ian Diamond was of great help in the search for the best models
and in the interpretation of two-way interactions. Formulas to derive the relevant coefficients and standard
errors of these interactions originated from him. Prof. G. W. Jones valuably commented this revised
version. Ms. Wendy Cosford patiently edited both the earlier and revised versions.
References
Alba, Richard D. (1987), 'Interpreting the Parameters of Log-Linear
Mode Is', Sociological Methods and Research, 13(l):77-107.
Antonovsky, Aaron and Judith Bernstein (1977), 'Social Class and
Infant Mortality', Social Science and Medicine, 2:453-470 .
Ballweg, John A. and Imelda Pagtolun-an (1988), 'Parental
Underinvestment :A Link in the Fertility-Mortality Continuum',
Paper presented at the Seminar on Mortality Transition in South
and East Asia, 29 August-2 September 1988, Beijing, sponsored
by the International Union for the Scientific Study of
Population and Institute of Population Research,
People's University of China.
Boulier, Bryan L. and Vicente B. Paqueo (1988), 'On the Theory and
Measurement of the Determinants of Mortality' , Demography,
25(2): 249-263.
Cabigon, Josefina V. (1982), 'The 1979 Mortality Survey in Southern
Mindanao', Final Report, Manila: UPPI, Mimeo.
Cabigon, Josefina V. (1990), 'Philippine Mortality in Changing Times' ,
Unpublished Ph.D. thesis, Australian National University,
Canberra.
Caldwell, John C. (1979), 'Education as a Factor in Mortality Decline:
An Examination of Nigerian Data, Population Studies,
33(3):395-413.
Caldwell J. C. and Peter McDonald (1981), 'Influence of Maternal
Education on Infant and Child Mortality: Levels and Causes',
in IUSSP, International Population Conference, Manila, 1981,
2:79-96, Liege: Ordina Press.
Casterline, John B., Elizabeth C. Cooksey and Abdel Fattah E. Ismail
(1989), 'Household Income and Child Survival in Egypt',
Demography, 26(l):15-35.
Cleland, John G. and Zeba A. Sathar (1984) 'The Effect of Birth
Spacing on Childhood Mortality in Pakistan', Population Studies,
38:401-418.
17
Clogg, Clifford C. and Scott R. Eliason (1987), 'Some Common Problems
in Log-Linear Analysis', Sociological Methods and Research,
16(l):8-44.
Concepcion, Mercedes B. (1982), 'Factors in Decline of Mortality in
the Philippines, 1950-1975', in WHO/ESCAP, Mortality in South and
East Asia: A Review ofChanging Trends and Patterns, 1950-1975,
Report and Selected Papers Presented at Joint WHO/ESCAP
Meeting held in Manila, 1-5 December, Manila:
Lyceum Press, Inc., pp. 327-354.
Concepcion, Mercedes B. and Josefina V. Cabigon (1982), 'Mortality
in Central Luzon and Metro Manila: the 1979 Mortality
Survey Final Report', Manila: UPPI, Mimeo.
Concepcion, Mercedes B. and Peter C. Smith (now Xenos)
(1977),
The Demographic Situation in the Philippines: An Assessment in 1977,
Papers of the East-West Population Institute, No. 44,
Honolulu: East-West Center.
Cox, D.R. and E. J. Snell (1981) , Applied Statistics: Principles and Examples,
London: Chapman and Hall.
Cramer, James C. (1987), 'Social Factors and Infant Mortality:
Identifying High-Risk Groups and Proximate Causes',
Demography, 24 (3) : 299-322.
DaVanzo, Julie, W.P. Butz and J.P. Habicht (1983), 'How Biological
Behavioural Influences on Mortality in Malaysia Vary
During the First Year of Life', Population Studies, 37:381-402.
Department of Health (1987a), 'Five-year Acceleration of Expanded
Program on Immunization in the Philippines, 1987-1991',
Manila: Department of Health.
Department of Health (1987b), 'Control of Diarrheal Diseases, 19881989 Plan of Operations', Manila: Department of Health.
Department of Health (1987c), 'Acute Respiratory Infection, 1988-1989
Plan of Operations', Manila: Department of Health.
De Sweemer, Cecile (1984), 'The Influence of Child Spacing on Child
Survival' , Population Studies, 38(1) : 47-72.
Edouard, Lindsay (1981), 'The Dynamics of Perinatal Mortality:
Maternal Age, Parity and Legitimacy,
Scandinavian Journal of Social Medicine, 9:59-61.
Engracia, Luisa T. (1983), 'Infant Mortality and Health Services in
Rural Philippines', Paper presented at the 6th National
Population Welfare Congress, 17 November 1983, Manila.
Frenzen, Paul D. and Dennis P. Hogan (1982), 'The Impact of Class,
Education, and Health Care on Infant Mortality in a Developing
Society: the Case of Rural Thailand', Demography,
19(3):391-408.
Gortmaker, Steven L. (1979),
'Poverty and Infant Mortality in the
United States', American Sociological Review, 44:280-297.
Haines, Michael R. and Roger C. Avery (1982), 'Differential Infant
and Child Mortality in Costa Rica: 1968-1973', Population Studies,
36:31-43.
18
Herrin, Alejandro N. (1988), 'Population, Human Resources and
Development: A Research Report', Submitted to the UNFPA project
implemented by the University of the Philippines Center for
Integrated and Development Studies.
Hill, Hal (1986), 'The Philippine Economy in 1985: The Decline
Continues', in Lim Joo-Jock (ed.), Southeast Asian Affairs 1986,
Singapore: Institute of Southeast Asian Studies, pp. 239-257.
Hill, Hal (1988), 'The Philippine Economy Under Aquino: New Hopes,
Old Problems', Asian Survey, 28 (3) : 261-285.
Hobcraft, John N., John W. McDonald and Shea O. Rutstein (1983),
'Child-spacing Effects on Infant and Early Child Mortality',
Population Index, 49:585-618.
Hobcraft, John N., John W. McDonald and Shea O. Rutstein (1984),
'Socio-economic Factors in Infant and Child Mortality: A
Cross-National Comparison', Population Studies, 38 (2) : 193-223.
Hull, Terence H. and Bhakta Gubhaju (1986), 'Multivariate Analysis of
Infant and Child Mortality in Java and Bali', Journal of Biosocial
Science, 18:109-118.
Johnson, Nan and Merwyn R. Nelson (1984), 'Housing Quality and
Child Mortality in the Rural Philippines',
Journal of Biosocial Science, 16:531-540.
Kiely, John L., Nigel Paneth and Mervyn Susser (1986), 'An Assessment
of the Effects of Maternal Age and Parity in Different Components
of Perinatal Mortality', American Journal of Epidemiology,
123(3): 444-454.
Knodel, J. and A. Hermalin (1984), 'Effects of Birth Rank,
Maternal age. Birth Interval, and Sibship Size on Infant and
Child Mortality: Evidence from 18th and 19th Century
Reproductive Histories', American Journal of Public Health,
74 (10) :1098-1106.
Knoke, David and Peter J. Burke (1981) , Log-Linear Models,
Beverly Hills: Sage Publications.
Lee, Eun Sul, Ronald N. Forthofer and Ronald J. Lorimor (1986),
'Analysis of Complex Survey Data',
Sociological Methods and Research, 15(1-2): 69-100.
Long, J. Scott (1984), 'Estimable Functions in Log-Linear Models',
Sociological Methods and Research, 12(4): 399-432.
Long, J. Scott (1987), 'Introduction to Common Problems in
Quantitative Social Research',
Sociological Methods and Research, 16(1): 3 - 7 .
Madigan, Francis C. (1977), 'Transition from the Demographic
Transition: Perspective from the Philippines', in
John C. Caldwell (ed.) , The Persistence of High Fertility:
Population Prospects in the Third World, Family and Fertility
Change Series 1, Part 2, Canberra: The Australian National
University, pp. 765-802.
Madigan, Francis C. (n.d.), 'Infant Mortality by Socioeconomic
Variables, Philippines 1983', in M.B. Concepcion (ed.),
Demography at the Crossroads (in press).
19
Martin, Linda G., J. Trussell, F.R. Salvail and N.H. Shah (1983),
'Covariates of Child Mortality in the Philippines, Indonesia
and Pakistan: An Analysis Based on Hazard Models',
Population Studies, 27 (3) : 417-432 .
McCullagh, P. and J. A. Nelder, F.R.S.
London: Chapman and Hall.
(1983) , Generalized Linear Models,
Merrick, Thomas W. (1985), 'The Effect of Piped Water on Early
Childhood Mortality in Urban Brazil, 1970 to 1976',
Demography, 22 (1) : 1-23 .
Palloni, Alberto (1989), 'Effects of Inter-birth Intervals on Infant
and Early Childhood Mortality', in L. Ruzicka, G. Wunsch and
P. Kane (eds.), Differential Mortality: Methodological Issues
and Bio social Factors, Oxford: Clarendon Press, pp. 163-188.
Palloni, Alberto and Marta Tienda (1986), 'The Effects of
Breastfeeding and Pace of Childbearing on Mortality at
Early Ages', Demography, 23(1): 31-52.
Payne, C.D. (ed.) (1985), The GLIM System Release 3,77, Generalized
Linear Interactive Modelling Manual, Oxford: Numerical
Algorithms Group.
Pharoah, P.O.D. (1976), 'International Comparisons of Perinatal and
Infant Mortality Rates' , Proceedings of the Royal Society
of Medicine, 69:335-338.
Pharoah, P.O.D. and J.N. Morris (1979),
Epidemiologic Reviews, 1:170-183.
'Postneonatal mortality',
Reyes, Fiorentina (1981), 'Evaluation of the Republic of the
Philippines Fertility Survey, 1978', WFS Scientific Reports,
No. 19. London: World Fertility Survey.
Rutstein, Shea O. (1983), 'Infant and Child Mortality: Levels and
Demographic Differentials, World Fertility Survey
Comparative Studies, No. 24, London: World Fertility Survey.
Schultz, T. P. (1980), 'Interpretation of Relations among Mortality,
Economics of the Household, and Health Environment',
Proceedings of the Meeting on Socioeconomic Determinants and
Consequences of Mortality, Mexico City, June 1979,
New York: United Nations, pp. 382-422.
Shapiro, Sam, Edward R. Schlesinger and Robert E.L. Nesbitt, Jr.
(1968), Infant, Perinatal, Maternal, and Childhood,
Mortality in the United States, Cambridge, Mass, :
Harvard University Press.
Swan, A. V. (1985), in C. D. Payne (ed.), The GLIM System Release 3.77,
Generalized Linear Interactive Modelling Manual, Oxford: Numerical
Algorithms Group.
“
Trussell, James and Charles Hammerslough
(1983),r'A Hazards- Model
Analysis of the Covariates of Infant and Child Mortality
in Sri Lanka', Demography, 20(l):l-26.
United Nations (1982), Levels and Trends of Mortality Since 1950,
New York: United Nations Department of International
Economic and Social Affairs.
Victors, Cesar G., Peter G. Smith and J. Patrick Vaughan (1986),
20
'Social and Environmental Influences on Child Mortality in
Brazil: Logistic Regression Analysis of Data From Census
Files' , Journal of Biosocial Science, 18:87-101.
Yudkin, P.L. and M. Baras (1983), 'A New Approach to Assessing the
Effect of Birth Order on the Outcome of Pregnancy',
Journal of Biosocial Science, 15:307-316.
Zablan, Zelda C. (1978), 'Trends and Differentials in Mortality', in
Population of the Philippines, Country Monograph
Series No. 5, Bangkok: United Nations ESCAP, pp. 99-116.
Zablan, Zelda C. (1983), 'Trends and Differentials in Mortality', in
Mercedes B. Concepcion (ed.). Population of the Philippines:
Current Perspectives and Future Prospects,
Pasig: National Economic and Development Authority.
21
Appendix
Derivation of the deviance and its corresponding
degrees of freedom of a given model
Given individual files that are too large because of many variables to consider, creation of a condensed
file is necessary for GLIM to run successfully. T. W. Pullum’s (personal communication. May 1, 1989)
GLIMTAB .FOR FORTRAN program constructs a condensed file in which the number of cases is equal to
the number of distinct combinations of the predictor variables. More simply, this program produces the
accumulated number of cases and amount of exposure in each cell. The deviances (G2C) and their
corresponding degrees of freedom (dfc) calculated from the condensed file served as basis for the search for
the optimal models in the present study.
To derive the deviance of a given model (G2m) and its corresponding degrees of freedom (dfm)
necessitates calculation of the baseline or null model deviance (G2q) from the individual file and its
corresponding degrees of freedom (dfg) which is simply the total number of cases minus one.
T. W. Pullum’s (personal communication. May 9, 1989) DEVBASE.FOR FORTRAN program calculates
G20 and df0 according to the formulas of McCullagh and Nelder (1983:25).
There is a simple relationship between a GLIM run on the individual file and a GLIM run on the
condensed (c) or aggregated file. That is.
G2c0 - G2cm = G20 - G2m-
Therefore,
G2m = G2o-(G2cO-G2cm,),and
:
dfm = dfo-(<ifco-dfcm)
where:
G2’c0
, = deviance for the null model based on the condensed file
°£ G2co
G2c:n = deviance for the more complex model based on the condensed file
dfcm = dfofG2cm.
22
Table 1: Demographic, socioeconomic and health related
differentials in infant and child mortality,
Philippines: all children, 1983 NDS
Mortality rate per 1000 births.
Variables
Neonatal
Post-neonatal
Infant
Child
45(9504)
45 (15255)
55(18220)
21 (8702
28 (13800)
36(16103)
Demographic
Birth order
1
2-3
4+
26(9504)
21 (15255)
22 (18220)
20 (9242)
24 (14897)
34 (17769)
Maternal age at birth of child (years)
<20
20-34
35+
31(5269)
21(33116)
22 (4593)
33(5099)
26(32326)
32 (4482)
63 (5269)
46(33116)
53(4593)
36(4766)
29 (29821)
27(3947)
28 (27723)
27(14185)
51(28473)
46(14506)
33(25537)
23(13367)
Socioeconomic
Residence
Rural
Urban
24 (28473)
20 (14506)
Presence of electricity in the household
Without 23(21963)
With
22(21016)
31(21405)
24 (20503)
53(21963)
45(21016)
38(19599)
21 (19005)
Average household monthly income
<P1000
P1000+
24 (32350)
18 (10629)
30 (31493)
19(10415)
53(32350)
37(10629)
35(28868)
14 (9737)
31(11932)
20(17531)
20(9152)
16(4363)
40(11539)
25(17150)
22(8939)
14 (4279)
70 (11932)
44 (17531)
42(9152)
30 (4363)
48 (10688)
30(15889)
15 (8160)
9(3867)
Education
<Elem.
Elem.
H Sch.
CO11.+
23
Table 1. Demographic, socioeconomic, and health related
differentials in infant and child mortality,
Philippines: all children, 1983 NDS (cont'n)
Mortality rate per 1000 births
Variables
Neonatal Post-neonatal
Infant
Child
Health Related
Source of drinking water
Unsafe
Safe
23(23858)
22(19121)
27(23265)
27 (18643)
49(23858)
49(19121)
32 (21435)
26(17169)
37 (8971)
26(24639)
20(8297)
62(9227)
48(25296)
38 (8456)
50 (8052)
28 (22825)
12(7727)
31(26590)
21 (15317)
53(27300)
42(15679)
37(24333)
17 (14272)
27 (10348)
28 (31560)
49(10621)
49(32358)
28 (9411)
30 (29192)
Presence of toilet
None
Outside
Inside
25(9227)
23(25296)
18 (8456)
Housing quality
Inadequate 23(27300)
Adequate
21(15679)
Household composition
Extended
Nuclear
23(10621)
22(32358)
( ) N of births
24
Table 2: Models that describe the relationships between
neonatal, post-neonatal, infant and child
mortality and demographic, socioeconomic and
health related-environmental factors: all children
Mortality
indicator/
model type
Model
Deviance
df
11668.10
11464.50
42470
42461
Neonatal
Null
Main effects P+ED+BO+M
Optimal
Main effects +
P.ED.M
11417.30
42433
Post-neonatal
Null
11036.38
41408
Main effects P+BO+IN+ED+BI+M4-H+T
10621.88
Infant
Null
Main effects P+BO+IN+ED+BI+M+T+W
Optimal
9
203.60
★
47.20
28
41394
414.50
14
21356.49
20847.49
42470
42456
★
509.00
14
20715.79
42433
131.70
23
11320.76
10789.46
38107
38093
*
531.30
14
Main effects+P.BO+P.ED+
IN.BO+IN.T.ED
10716.66
38066
72.80
Main effects+P.M+
P.W+ED.T.W
Child
Null
Main effects IN+T+ED+P+BO+BI+H+M
Optimal
Difference from
the simpler model
Deviance
df
★
27
Notes: Null=baseline grand mean model or a model with no covariates
in which a single overall value of the rate is assumed to apply to
entire population; optimal model=most adequate description of the
relationships between the response variable and its covariates;
main effects model=adequate description of the relationships between
the response variable and its covariates excluding significant
interaction terms between the covariates; P=length of preceding
birth interval; BO= birth order; M=age of mother at birth of child;
IN=average household monthly income; ED=mother's education;
H=housing quality; T=toilet facility; W=source of drinking water;
BI=Period of birth; A dot means interaction - two-way
and three-way.
The deviance measures how closely the model fits the
data and it is distributed asymptotically as chi-square under the
null hypothesis that the model used and the underlying assumptions
are correct. The degrees of freedom (d.f.) indicate how much
information is available for estimating the 'background noise'
with the residual variance.
★
Statistically significant at 0.025 or lower level.
Three-way interaction, but not two-way, among IN, ED and M is
statistically significant at 0.05 level; however the estimated
interaction effects were Small relative to the main effects.
Hence on the basis of parsimony (a model with minimum number of
parameters), the main effects model is the optimal model.
25
Table 3: Parameter estimates and relative rates from the main
effects and optimal models for neonatal mortality
Model/Parameter
Optimal
Main effects
Variable
Estimate Standard Relative
error
rate
Constant or
grand mean
0.14
-2.05
Estimate Standard Relative
error
rate
0.129
-1.99
1.000
0.00
0.28
0.137
A. No interactions
Birth order
1
0.00
2-3
-0.44
1.000
*
★
0.12
0.644
-0.40
0.14
0.670
0.12
0.631
-0.45
0.14
0.638
★
4 and over
-0.46
Maternal age at childbirth (in years)
<20
1.000
0.00
★
20-34
35 & over
0.10
0.15
-0.45
-0.15
Mother's education
Prim. & below 0.00
0.638
0.861
1.000
★
Elementary
-0.41
0.08
0.664
0.09
0.794
0.13
0.698
★
High School
-0.23
*
College +
-0.36
26
Table 3:
Parameter estimates from optimal model with significant
interactions for neonatal mortality (continuation)
Preceding birth
Education of Mother
birth interval/---------maternal Primary & below Elementary
High School
College*
age at
--------------childbirth Estimate S.E. Estimate
1
S.E. Estimate S.E. Estimate S.E.
B. With interactions
<18
<20
0.00
20-34
-0.37
-0.32
0.40
-0.61
0.56
-3.79
6.73
0.27
-1.07
0.77
-0.90
1.03
-1.21
11.67
35 & over -0.63
0.38
-0.31
0.94
-0.60
1.38
-4.16
14.14
18-30
<20
-0.71
0.47
-2.11
1.24
-0.31
1.41
-0.62
11.73
20-34
-1.41
0.89
-1.75
2.08
-1.80
2.37
-1.59
19.16
35 & over -1.03
1.06
-2.17
2.39
-1.08
2.90
0.07
21.59
0.32
-1.09
0.82
-1.18
1.09
-0.27
11.69
-1.33
0.64
-1.76
1.46
-1.30
1.88
-1.62
19.10
35 & over -1.08
0.86
-1.23
1.79
-1.08
2.57
-1.69
22.94
31 & over
<20
*
-1.05
★
20-34
Note:
The relative rate of the constant is the reference rate.
Preceding birth interval is a significant control.
Statistically significantly different from zero.
27
Table 4: Parameter estimates and relative rates from the main
effects or optimal model for post-neonatal mortality
Parameter
Variable
Estimate
Standard error
Relative rate
Constant or
grand mean
-1.67
0.16
0.188
Birth order
1
0.00
1.000
*
2-3
4 and over
-0.29
0.11
Household income
<P1000
0.00
0.12
0.12
0.748
1.116
1.000
★
P1000+
Mother's education
Prim. & below
-0.21
0.08
0.811
1.000
0.00
*
Elementary
-0.37
0.07
0.691
0.09
0.684
0.14
0.527
★
High School
-0.38
★
College +
-0.64
Maternal age at childbirth (in years)
<20
0.00
★
20-34
-0.36
0.10
-0.24
0.13
35 & over
Housing quality
Inadequate
0.00
1.000
0.698
0.786
1.000
*
Adequate
-0.16
Toilet facility
None
0.00
0.07
0.852
1.000
*
Outside
-0.24
0.07
0.787
Inside
-0.17
0.11
0.844
The relative rate of the constant is the reference rate.
Note:
Significant controls are preceding birth interval and time period.
★
Statistically significantly different from zero.
28
Table 5: Parameter estimates and relative rates from the main
effects and optimal models for infant mortality
Model/Parameter
Variable
Optimal
Main effects
Estimate Standard Relative
error
rate
Constant or
grand mean
A.
0.12
-1.20
Estimate Standard Relative
error
rate
0.301
-1.04
1.000
0.00
0.18
0.353
No interactions
Birth order
1
0.00
2-3
4 and over
0.08
-0.40
-0.13
1.000
★
★
0.09
0.670
0.878
0.10
0.10
-0.40
-0.14
0.670
0.869
Maternal age at childbirth (in years)
1.000
<20
0.00
20-34
35 & over
0.07
0.10
-0.27
-0.18
Household income
<P1000
0.00
0.763
0.835
1.000
0.00
P1000+
-0.20
0.06
Mother's education
Prim. & below 0.00
Elementary
0.819
1.000
-0.41
0.05
0.664
★
High School
-0.41
0.06
0.664
College +
-0.62
0.10
0.538
Drinking water source
Unsafe
0.00
★
0.12
Safe
1.000
0.04
Toilet facility
0.00
None
1.127
1.000
★
Outside
-0.13
0.05
0.878
0.08
0.795
★
Inside
-0.23
1.000
★
*
-0.21
0.06
0.811
29
Table 5:
Parameter estimates from optimal model with significant
interactions for infant mortality (continuation)
Interaction
term/variable
Variable/parameter
B. Predictors with 2-way interactions
Maternal
Preceding birth interval (months)
age at —
child
<18
18-30
31 & over
birth
Est. S.E. R.Rate Est. S.E. R.Rate Est. S.E. R.Rate
*
<20
1.000
0.00
20-34 -0.24 0.13 0.787
35+
-0.23 0.17 0.794
-1.05 0.19
★
-1.19 0.36
★
-1.14 0.43
★
-1.11 0.17 0.330
★
-1.43 0.31 0.239
★
-1.20 0.39 0.301
0.350
0.304
0.320
C. Predictors with significant 3-way interactions
Toilet
facility/
water
source
None
Unsafe
Education of mother
Primary & below Elementary High School College+
Est.
S.E.
Est.
S.E.
-0.57
0.13
-0.20
0.19 -1.86
0.93
0.35
-0.55
0.45 -0.11
1.70
Est.
*
★
0.00
S.E.
S.E.
Est.
★
Safe
-0.46
0.15
-0.44
Outside
Unsafe
-0.09
0.09
-0.60
0.27
-0.87
0.37 -1.19
1.64
Safe
-0.42
0.29
-0.71
0.64
-0.60
0.81 -0.63
2.83
Inside
Unsafe
-0.34
0.27
-1.17
0.60
-0.86
0.64 -0.93
1.71
Safe
-0.51
0.57
-0.62
1.14
-0.62
1.23 -1.06
2.95
★
*
The relative rate of the constant is the reference rate.
Note:
Significant controls are preceding birth interval and time period.
*
Statistically significantly different from zero.
Nearly statistically significantly different from zero.
30
Table 6: Infant mortality rate (per 1000 births) by
education and water(most detailed categorization): all children
Residence/
water
source
Education
Primary & below Elementary High School College+
All
Lake/river/stream 118 (501)
Spring
75(1842)
Rainwater
54(240)
Open well
71(2449)
Pump, shallow well 69(2892)
Artesian,deep well 71(2027)
Pipe water
52(1974)
51(354)
40(1483)
31(323)
47(2953)
35(5141)
44 (3322)
54(3951)
56(178)
34(535)
21(187)
35(882)
34 (2416)
53(1802)
45(3140)
56(18)
30(135)
43(47)
53(282)
32(1229)
47(1025)
57(2262)
29(35)
118 (17)
38(78)
64(47)
37(27)
34 (118)
26(39)
42(914)
11(449)
55(874)
31(422)
47 (2311)
30 (1655)
51(336)
40(1348)
33(276)
46(2672)
36(3911)
43(2297)
50 (1689)
63(142)
33(457)
19(160)
37 (764(
29(1503)
51(928)
37(829)
59(51)
49(123)
36(139)
19(155)
18 (986)
40 (866)
32(2042)
Urban
Lake/river/stream
Spring
85(130)
Rain water
39(51)
Open well
53(207)
Pump, shallow well 76(589)
Artesian,deep well 72(624
Pipe water
52(893)
Rural
Lake/river/stream 122(485)
Spring
75(1713)
Rainwater
58(189)
Open well
73(2242)
Pump, shallow well 67(2303)
Artesian well
71(1403)
Pipe water
52(1080)
( ) N of births
59(51)
38(105)
21(93)
7(115)
24(537)
50(444)
39(387)
31
Table 7: Parameter estimates and relative rates from the main
effects and optimal models for child mortality
Mode1/Parameter
Variable
Optimal
Main effects
Estimate Standard Relative
error
rate
Constant or
grand mean
A.
-2.73
0.15
0.065
0.12
1.000
0.980
0.12
1.377
Estimate Standard Relative
error
rate
-2.50
0.18
0.082
No interactions
Birth order
1
2-3
0.00
-0.02
★
4 and over
0.32
Maternal age at childbirth (in years)
1.000
<20
0.00
*
20-34
0.10
0.811
-0.21
1.000
0.00
-0.25
0.10
0.779
0.15
0.657
*
35 & over
0.14
-0.40
Household income
<P1000
0.00
0.670
-0.42
1.000
*
P1000+
0.10
-0.40
Mother's education
Prim. & below 0.00
0.670
1.000
*
Elementary
-0.29
0.06
0.748
0.11
0.477
0.19
0.440
*
High School
-0.74
*
College +
-0.82
Housing quality
Inadequate
0.00
1.000
Adequate
*
0.08
-0.34
Toilet facility
0.00
None
0.712
1.000
*
Outside
-0.45
0.06
0.638
0.13
0.445
★
Inside
-0.81
1.000
0.00
*
-0.33
0.08
0.719
32
Table 7:
Parameter estimates from optimal model with significant
interactions for child mortality (continuation)
Variable/parameter
Interaction term/
variable
B. Predictors with 2-way interactions
Household income
Birth
order
P1000 & over
<P1000
Estimate
1
0.00
2-3
-0.35
R.Rate
Estimate
S.E.
R.Rate
1.000
-0.52
0.37
0.594
0.17
0.705
-0.06
0.58
0.942
0.14
1.363
0.20
0.57
1.221
S.E.
★
4+
0.31
C. Predictors with 3-way interactions
Education of mother
Toilet
facility/
household Primary & below Elementary
High School
income
—’——--—— —--- ----Estimate S.E Estimate S.E Estimate S.E.
College+
Estimate S.E.
None
*
<P1000
0.00
P1000+
-0.52
-0.41
-0.47
0.16
-0.58
0.25
0.27
0.50
0.37
-1.31
0.78
-6.28
6.96
-2.45
1.67
0.10
-0.82
0.29
-1.62
0.42
-1.68
0.87
0.61
-1.20
1.34
-2.56 11.02
-2.90
2.91
0.71
-3.30
1.30
11.11
-3.49
3.34
Outside
*
*
*
<P1000
★
P1000+
-1.61
Inside
★
★
CP1000
-0.34
0.25
-0.93
0.55
-2.19
P1000+
-1.78
0.92
-2.21
1.82
-3.24
Note: The relative rate of the constant is the reference rate.
Significant controls are preceding birth interval and time period.
*
Statistically significantly different from zero.
Nearly statistically significantly different from zero.
■ '• .w •-
CHILD
SURVIVAL
Research Note Number
36CS
Date
18 April 1991
Division of Demography and Sociology
Research School of Social Sciences
The Australian National University
Canberra ACT, Australia
A Project Sponsored by The Ford Foundation
CHILD IMMUNIZATION IN BURUNDI AND ZIMBABWE
Christine McMurray
Graduate Program in Demography,
National Centre for Development Studies,
The Australian National University
and
Masauso M. Nzima
Graduate Program in Demography,
National Centre for Development Studies,
The Australian National University
Note:
Child Survival Research Notes are brief discussions
of issues of current relevance to researchers and
policy-makers concerned with problems of high infant
and child mortality in the world. The International
Population Dynamics Program, Division of Demography
and Sociology, The Australian National University
distributes these notes with their regular
Bibliographic Circular. Production of the Child
Survival Research Notes is made possible through a
grant from the Ford Foundation. Responsibility for
the content of Child Survival Research Notes rests
with the author(s) alone, and not the above-listed
organisations.
1990 - Celebrating Sixty Years of Higher Education in the ACT
ANU Open T>ay - Sunday 16 September 1990
1
Burundi and Zimbabwe have experienced different levels of political stability and
economic growth since attaining independence. Currently Burundi is among the poorest
countries in Africa while Zimbabwe is one of the more prosperous. 1 During the early
1980s both countries adopted the major World Health Organisation initiative, the
Expanded Programme of Immunization (EPI). This paper compares patterns of
immunization in the two countries, and their association with other aspects of child
healthcare.
THE EXPANDED PROGRAMME OF IMMUNIZATION (EPI)
The EPI focuses on immunization against six of the major vaccine-preventable childhood
diseases: tuberculosis, diphtheria, pertussis, tetanus, poliomyelitis and measles.
Henderson (1984: 3) estimates that together these diseases kill an average of five million
children each year and cripple or mentally retard many of those who survive. Measles
alone accounts for one third of all infant and child deaths in Burundi (UNESCO, 1988: 4).
Since immunization is generally very effective in preventing disease, comprehensive
immunization coverage against the six major diseases can do much to reduce infant and
child mortality. However, it should be remembered that immunization is not yet widely
available for one major cause of infant and child mortality, diarrhoeal disease. Moreover,
some other immunizeable diseases, such as influenzas, which at times may also present
a serious risk to children, are not currently specified as EPI targets.
The EPI promotes use of BCG, DPT, polio, measles and tetanus toxoid vaccines. The
first four of these are given to children, and the fifth vaccine, tetanus toxoid, is
administered to pregnant women so their antibodies can be transferred to the foetus to
protect against neonatal tetanus.
Table 1 shows the recommended immunization schedules for children in Burundi and
Zimbabwe. That for Zimbabwe approximates the EPI general recommendations of at or
near birth for BCG; six weeks and onwards with not less than 28 day intervals for three
doses of DPT and polio; and 9 months for measles (Henderson, 1984: 21). However,
the schedule for Burundi recommends longer courses of DPT and polio, with four rather
than three doses of each of these vaccines.
1 UNICEF (1990: 84) reports a 1987 per capita Gross Domestic
Product of only US $ 250 per annum for Burundi compared with
US $ 580 for Zimbabwe.
2
The authors are currently seeking more information on the reason for these additional
doses. As it is recommended that the fourth dose of DPT should be given "at some
stage" it is assumed to be a booster that may be given at a much later age. On the other
hand, the first dose of polio is recommended at birth, with the fourth dose at 14 weeks,
when most countries are giving only a third dose. The fourth dose could be a booster
containing all three strains of polio, or possibly a local epidemic or high risk conditions
require a repeat dose of one strain.
The EPI has boosted immunization programmes in participating countries primarily by
improving availability of vaccines and the efficiency of the "cold chain" for storage, and by
training and health promotion. UNICEF (1985: 144) says of Zimbabwe
"Vaccine shortages have become a feature of the past....Particularly important
was the establishment of outreach services enabling the programme to reach a
high proportion of eligible children and women..."
However, although in each country there have been substantial improvements in the
number of children protected by the major vaccines (UNICEF, 1985: 144, UNESCO,
1988: 7), neither country has achieved complete coverage of all eligible children. The
following analysis focuses on differences in coverage and timing of immunization in the
two countries.
THE DATA
This study is based on two Demographic and Health Surveys (DHS) conducted by the
Institute for Resource Development: the Enquete Demographique et de Sante au
Burundi 1987, (EDSB), and the Zimbabwe Family Health Survey II, 1988 (ZDHS). In each
country all women aged 15-49 were interviewed in a nationally representative, two-stage,
self-weighting sample of households. A total of 3970 women were interviewed in Burundi
and 4201 in Zimbabwe. The questionnaire was wide ranging and collected detailed
information on the healthcare given to children under age five. The present analysis uses
a subset of children aged sixty months or less to allow comparison of the care received
by individual children rather than the care given by mothers.
The datasets included 3885 children aged sixty months and under in Burundi and 3527
for Zimbabwe. Some are siblings with the same mothers. Fifty-one children whose
mother's ethnicity was classed as "white", "coloured" or "other" were excluded from the
Zimbabwe analysis, leaving a total of 3576. No children were excluded from the Burundi
subset on the basis of ethnicity as these data were not included in the EDSB.
3
The DHS collected data on immunization for all children under age sixty months in the
sample. However, if the mother did not show the interviewer a Child Health Card, which
contains information on the date and type of immunization given, she was asked only
whether the child had ever been vaccinated and no information was recorded on timing
and type of vaccinations for that child. Only 47 per cent of surveyed children aged five
years and under in Burundi and 73 per cent in Zimbabwe had a health card (ZDHS,
1989:85; EDSB, 1988: 70).
PROPORTIONS IMMUNIZED
Table 2 compares the results of small scale monitoring surveys of children aged 12-23
months carried out by the Zimbabwe Ministry of Health with the children of the same age
surveyed by ZDHS. It can be seen that in most areas the percentage who had received
all of the recommended vaccinations increased substantially between 1982 and 1984 or
1985, which reflects the impact of the EPI. In 1982 only 25 per cent of one year old
children outside the three main cities surveyed by the Ministry of Health (MOH) were fully
immunized, compared with 42 per cent in 1984 (UNICEF, 1985: 3). MOH data for two
other major cities, Bulawayo and Chitungwiza, show slightly higher rates of coverage in
both 1982 and 1984-5 than for Harare. The report does not offer an explanation for this
difference, although possibly it could be because Harare is larger and more
heterogeneous.
The ZDHS data for 1988 refer to the country as a whole as it is not known if the
boundaries for the MOH surveys were comparable with DHS regions. However, it is
clear that there has been a continuing increase in proportions immunized. The Report of
the 1988 ZDHS compared children aged 12 to 23 months for whom a health card was
produced with children of the same age surveyed by MOH (ZDHS, 1989: 88). Seventy
eight per cent of children in this age group in ZDHS had a health card, but a concurrent
survey by MOH found 94 per cent had a health card. ZDHS attributes this to differences
in sampling methods (ZDHS, 1989: 88). However, of the children with cards, ZDHS
reports that 86 per cent were fully immunized compared with only 80 per cent in the MOH
survey. Our analysis of the cleaned ZDHS dataset found 82 per cent of children aged 12
- 23 months had a healthcard, of whom 86 per cent were fully immunized. Despite these
variations between surveys, there appears to be a marked progressive improvement in
proportions with a health card and proportions fully covered from 1982 to 1988.
Table 2 also shows a substantial decline in the drop-out rates in Harare and Zimbabwe
outside the three main cities. It is not stated how many cases are used as a basis for
these findings, but the numbers are probably small, comparable to those shown for
4
immunization coverage. Analysis of the ZDHS showed an extremely low drop-out rate of
only 2 per cent between first and second doses of DPT and polio and only 6 per cent
between first and third doses. Clearly the EPI is not only successful in recruiting
acceptors but also in providing follow up services.
Comparative data for Burundi were not available to the authors at the time of writing, but
immunization coverage has apparently improved since the introduction of the EPI.
UNESCO (1988: 7) reports that by December 1987 55 per cent of children were fully
immunized, in line with targets. This source also states that by 1990 there is expected to
be complete immunization coverage of at least 80 per cent of children under one year of
age (UNESCO, 1988: 11).
Increases in immunization coverage coincide with declining infant and child mortality
rates, particularly in Burundi. The EDSB data indicate that the infant mortality rate
declined from 100 per thousand live births in 1972-76 to 75 per thousand in 1982-86, and
the under five mortality rate from 224 to 152 in the same period (EDSB, 1988: 63).
Although during the 1970s infant and child mortality rates in Zimbabwe were already
substantially lower than those in Burundi, they too appear to have declined during the
1980s. ZDHS data show infant mortality levels hovering around 53 per 1000 in
Zimbabwe in both 1973-77 and 1983-88, although it is observed that the earlier figure is
probably an underestimate and a steady decline is more probable (ZDHS, 1989: 78).
The under five mortality rate is estimated to have declined from 92 in 1973-77 to 75 in
1983-88.
It should be noted that these estimates are somewhat lower than UNICEF estimates of
111 per thousand for Burundi and 71 for Zimbabwe in 1988 (UNICEF: 1990: 76). The
latter is consistent with an infant mortality rate of 79 derived from the 1984 Zimbabwe
Reproductive Health Survey (1985: 73). UNICEF (1985: 12) comments that there is
systematic under-reporting of infant and child deaths in Zimbabwe because people do
not like to speak of them. Thus, while there seems little doubt that the adoption of the
EPI has coincided with a downward trend in infant and child mortality in both countries,
the present levels could be higher than those reported by DHS.
The Burundi recommendation of four doses each of DPT and polio reduces the
probability of a Burundi child completing the schedule compared with a child from
Zimbabwe, who requires only three doses of each to be regarded as fully immunized. As
mentioned above, information is not yet available to the authors as to why the Burundi
5
recommendation was made, the effectiveness of three dose courses compared with four
dose courses in that country or the optimum age for a fourth dose of DPT. This study
will therefore follow the example of the EDSB report (1987: 71, Table 6.8) and use three
doses as the basis for estimating immunization coverage.
Table 3 presents the raw proportion of all surveyed children who had received each of
the vaccinations recommended in the country schedule. About 15 per cent of children in
each sample were aged less than 9 months so not eligible for all vaccinations, but all
children were eligible for BCG. Eligibility will be considered in more detail in age
standardized tables presented below. Table 3 shows that in Burundi between 7 and 39
per cent of all surveyed children had received each of the 10 recommended
vaccinations. The lowest proportion is for a fourth dose of DPT. Although more than
twice as many children had received a fourth dose of polio, the percentage is still well
below that for the first three doses. If fourth doses are not considered at least 25 per
cent of children have received each of the remaining eight vaccinations. Of 1832 who did
not have a health card more than half, or a further 24 per cent of all children surveyed,
were reported as having received a vaccination at some time. Although it is not known
what this vaccination was, this data increases the possible range of the proportion
immunized.
Dunn and Yumkella (1990: 99) used two sets of assumptions about rural children aged
12-23 months in the EDSB for whom a health card was not produced at the interview but
who were reported as having ever had a vaccination. They concluded that the
proportions immunized against measles could range from 47 if none of those who did
not produce a card had been immunized to 60 per cent if all had been immunized.
The Interviewer's Manual (1987: 62) instructs interviewers to encourage respondents to
look for the health card and to give them time to do this by being patient and not
conveying the impression of being in a hurry. It thus seems that the most likely reason
for a mother being unable to produce a health card at the time of the DHS interview is
that it was lost. This implies that it may not have been considered to be very important
and, in the case of younger children, probably had not been used very often. Assuming
that the mother had answered the question correctly, it is thought that children for whom
a card was not produced are most likely to have received only BCG, and are unlikely to
have been fully immunized.
It can be seen from Table 3 that in Zimbabwe the proportions of all surveyed children
immunized were much higher than in Burundi. Fifty per cent or more had received every
6
immunization, with the figure as high as 63 per cent for BCG which is given at or near
birth. Of the 27 per cent who did not have a health card 682 were reported as having
received a vaccination at some time in their life, which is a further 19 per cent. However,
as mentioned above, it is unlikely that these children had received more than BCG.
Figure 1 compares the proportions who had received eight vaccinations, i.e. BCG, three
doses each of DPT and polio and measles, using as the base population only the 47 per
cent of surveyed children in Burundi and 73 per cent in Zimbabwe who had a health
card. It can be seen that although the percentages are roughly equal for immunizations
given to very young children, the drop out rate at older ages is greater in Burundi. This
suggests a less effective immunization campaign and probably also reflects poorer
access to and provision of health services in Burundi.
TIMING OF IMMUNIZATION
In order to have her child immunized a mother is usually expected to take it to a medical
facility such as a clinic. Some regions of Burundi and Zimbabwe may be served only by
mobile clinics which visit at weekly or monthly intervals. Even where there are permanent
health centres vaccinations are commonly offered only at certain times or on certain
days, in both developed and developing countries. This may create problems for
mothers who rely on public transport, or need to fit in with family farming activities or
whose children are sick on a given day. Such difficulties can lead children to receive all
or some of their vaccinations at ages and intervals which differ from those recommended
in the schedules presented in Table 1.
One limitation of the Burundi dataset available for this study is that it does not support
calculation of the exact age at which a vaccination was received. Although the day,
month and year of birth, as well as the date of vaccination, is usually recorded on child
health cards, the dataset includes only the child's month and year of birth as reported by
the mother, and not the date on the health card. The Zimbabwe dataset includes both
the exact date of birth as recorded on the health card and the date of vaccination for
most children with a health card, but the authors have are still waiting on information on
the accuracy of reporting of day of birth. An earlier question on date of birth gives only
the month and year of birth reported by the mother.
The day of immunization was not considered in the following tables so that the patterns
for the two countries can be compared. Month and year of birth as reported by the
mother was subtracted from the month and year of vaccination to determine the child's
7
age when vaccinated. Accordingly, all age estimates in the following tables could be plus
or minus 30 days.
Figure 2 plots the age at which vaccinations were received. Although in both countries
the data take the form of a normal or skewed distribution, there is substantial heaping on
the recommended age. This is particularly true of Zimbabwe where large proportions
received vaccinations at exactly the recommended age, even for immunizations requiring
three doses. As noted above, the numbers receiving fourth doses of DPT and polio in
Burundi are small, and in the absence of more information about these doses it is not
possible to draw any firm conclusions about them from this figure. For the remaining
eight vaccinations the pattern is much like Zimbabwe only with more scatter about the
recommended age. The figure also shows that slightly higher percentages received
vaccinations at later ages in Burundi. This suggests that the popularity of immunization
may be increasing, as mothers could be bringing in older children for immunizations they
may have missed. In both countries there is a wider spread for repeat doses of DPT and
Polio than for initial doses of all vaccines. This pattern probably applies in most countries
and is almost certainly due to universal 'human nature' rather than to any local
characteristics.
COMPLETENESS OF IMMUNIZATION
Of the four childhood vaccines, BCG and measles confer immunity with a single dose but
polio and DPT require two or three follow-up doses. Because immunization schedules
take time to complete and because of variations in timing it is difficult to compare the
immunization status of very young children with that of older children. One approach,
used by EDSB (1987) and Dunn and Yumkella (1990), is to consider only children aged
12 to 23 months.
The present paper takes a different approach and gives each child a percentage score
according to the proportion of appropriate vaccinations it has received. A vaccination is
considered appropriate if it is recommended for a child of that age or younger. For
example, as shown in Table 1, it is recommended that a Zimbabwe child aged four
months should have received one BCG, one DPT and one polio vaccination. A child of
that age who had received all three would be given a score of 100 per cent. A child of
the same age who had received only two vaccinations would be scored as 67 per cent.
A Zimbabwe child aged eight months is eligible to receive BCG plus three each of DPT
and polio. If it had received four of these its score would be 57 per cent. This is less
than the younger child who had received only two vaccinations, because it represents a
smaller proportion of the vaccinations for which the older child was eligible.
8
The calculation of immunization cover in the following tables is based on eight
vaccinations for both countries, and excludes data on fourth doses of DPT and polio in
Burundi because of uncertainty about their optimum timing and impact. Stoekel (1985:
119) writes
"It is now generally recognised that two doses of a good tetanus and diphtheria
toxoid are sufficient, especially if given with an interval of two months or more.
The only need is vaccine of appropriate quality".
Among the advantages of schedules requiring fewer vaccinations he lists convenience,
reduced costs, efficiency of protection with more potent vaccines and a reduction in the
drop out rate. Presumably the Burundi recommendation for a fourth dose derives from
use of inferior vaccines in the past, and it is possible that they may no longer be essential
to give protection.
Most of the following tables include all surveyed children, not just those whose health
card was seen. As such they represent the lower boundary for estimates of
immunization coverage for the entire sample. As stated above, the additional group who
claimed to have a health card which was not seen by the interviewer is thought to be
unlikely to have received the full course of recommended vaccinations.
Table 4 compares the immunization cover of all surveyed children according to age
group. It can be seen that the patterns are similar in the two countries, although in
Burundi a much higher proportion of children had not received any vaccinations at all.
Both countries also show a drop off in the numbers returning for repeat doses of DPT
and polio, especially Burundi.
The data include both children born before the EPI was fully implemented and those bom
after it was well established. The effectiveness of the programme is reflected in the
higher proportions fully covered in the 10-18 month age group than in the oldest age
group in both countries. Lower proportions in the youngest age group can be attributed
to delayed immunization. Whereas in Zimbabwe 49 per cent of surveyed children aged
37 to 60 months had no immunization cover at all, only 21 per cent of those aged 18
months or less had no cover. The picture is similar for Burundi, where 76 per cent of
children aged 37 to 60 months had no cover compared with only 39 per cent aged 10 to
18 months. Surprisingly, in this country the proportion with no cover in the age group 0
to 9 months rises to 52 per cent. This could possibly be due to a faltering in the
programme, or more probably, a tendency to deliver immunizations after the
recommended age.
9
It is interesting to note in Table 4 that the largest proportions in both countries had either
no cover or around 100 per cent, with few in the middle categories. This suggests that
continuation rates are reasonably high once children enter the immunization programme.
Nonetheless it would appear that there is scope for increasing immunization coverage if
more effective vaccines are made available and complete immunity from the major
diseases could be achieved with six vaccinations as suggested by Stoekel (1985).
Table 5 looks at the age at receiving each of eight vaccinations. In this table children
were allotted a score for each vaccination only if it was received at the age
recommended by the EPI. The same schedule was used for both Burundi and
Zimbabwe, for reasons discussed above. If a vaccination was received before or after
the recommended age a zero was allotted. As in Table 4, scores for all vaccinations
were totalled and divided by the maximum score a child of that age could receive if it
were up to date on the immunization schedule recommended for its country.
The table shows that except for BCG, which is usually given at birth, there appears to be
a considerable discrepancy between the actual timing of immunizations and the
recommended schedule. In Burundi 58 per cent of children aged 0-3 months received
their immunizations at the recommended time, but this falls to less than 10 per cent for all
other age groups. Similarly, in Zimbabwe the figure falls from 69 per cent to around 40
per cent or less for all other age groups. In contrast to Table 4, the largest proportions in
Table 5 are in the middle categories rather than at the two extremes. This indicates a
moderate amount of variation from the recommended schedule in both countries.
It is not obvious from the literature exactly how much the timing of immunizations can be
varied without impairing their effectiveness. Although the EPI objective is to complete
immunization in the first year of life, schedules which take longer may still be effective.
Some elasticity may be permissible in the interval between doses of DPT and polio
without impairing their effectiveness. For example, Muller et al. (1984: 95-107) explored
the epidemiology of pertussis in Machakos, Kenya, and concluded that two doses of
DPT given six months apart provided adequate protection, although three doses at
shorter intervals is preferable.
IMMUNIZATION AND SOCIO-ECONOMIC CHARACTERISTICS
The success of an immunization campaign depends on its widespread acceptance in the
community. The process of injecting vaccines usually upsets the child to some extent,
and often results in minor symptoms of a reaction against the serum. In a very few cases
10
vaccination may result in serious complications or even death (see for example Miller et
al., 1982). It is most important, therefore, that mothers are aware that the benefits of
immunization outweigh the disadvantages and the risk involved.
Mothers who have received some schooling are usually thought to be more receptive to
new ideas and better able to make decisions which involve some risk. Accordingly, more
children with educated mothers are expected to be immunized than children of
uneducated mothers. Table 6 looks at immunization coverage of all surveyed children
according to mother's education. This table uses the same index as Table 4, basing the
score on whether the child ever received each vaccination and including both children
who did and did not show a health card. Surprisingly, the table shows no consistent
difference between uneducated and educated mothers in patterns of immunization,
especially in Zimbabwe. The biggest discrepancy was in Burundi for children aged 0 to 9
months, where more than twice as many mothers with no education had not immunized
their children compared with mothers with higher education. However, as only five per
cent of all Burundi children in the sample had mothers with higher education, and only 20
per cent had mothers with primary education, further analysis with more cases would be
needed to determine whether the difference is due to education per se or to some other
associated factor such as higher socio economic status.
The lack of a consistent pattern across education groups in the rest of the table suggests
that in both countries the immunization programme is being promoted so as to reach all
education levels. Nonetheless, the difference in immunization coverage between the two
countries could be in part due to a 'mass education' effect in Zimbabwe. Caldwell (1989:
106-7) presents evidence that mass education leads to a change in the whole society's
attitude towards health and healthcare, and suggests that urbanization, as well as
education, may affect attitudes to childcare and utilisation of modern facilities. In these
circumstances different educational levels of individual mothers within a country might be
expected to be less important than differences between countries. Streatfield et al.
(1986: 5) describe the different attitudes of educated women to health services, and note
a tendency for them to respond to unsatisfactory treatment by complaining and
demanding better service rather than by failing to return for further treatment. The effect
of mass education may thus be to permeate attitudinal changes throughout the entire
society, regardless of differences in educational levels between individual mothers.
The pattern for immunization by place of residence shown in Table 7 similarly shows little
variation between categories. However, it seems there may be a tendency for
11
vaccinations to be received later in rural areas in both countries. This is indicated by the
higher proportion of children aged 0 to 9 months with no cover.
Immunization awareness and acceptance is generally developed in a population through
education and advertising campaigns. It might thus be assumed that mothers who live in
a household which owns a radio would be more likely to immunize their children. Table 8
compares immunization status for the two groups and shows that there is not much
difference according to radio ownership in either country. In Zimbabwe 37 per cent
without a radio had no cover compared with 34 per cent with a radio. In Burundi the
difference was greater, with 61 per cent of radio owners having no cover compared with
54 per cent of those without a radio. However, in neither country was the difference so
much as to suggest that radio ownership has a noticeable impact on immunization.
Rather it would seem that the immunization campaign is reaching respondents through
other channels, or that possibly community radios are available to spread health
messages. If, on the other hand, radio ownership is viewed as a proxy for economic
status, it would seem that the EPI is reaching all socio-economic groups.
MATERNAL IMMUNIZATION AND ORAL REHYDRATION THERAPY
Table 9 looks at child immunization coverage according to whether the mother received
an ante-natal vaccination with tetanus toxoid. It was hypothesized that if the mother was
introduced to the immunization programme and vaccinated herself before the child was
born she would be more likely to continue in the programme and have a fully immunized
child. As expected, the proportions with no cover among those whose mothers were
immunized are consistently lower than among those whose mothers were not
immunized, in both countries and for children of all ages. Similarly, higher proportions of
the children whose mothers were immunized had received 76-100 per cent of the
vaccinations for which they were eligible. This suggests that initial contact with the
mother is important to promote continuation to a programme of child immunization.
Table 10 relates mother's knowledge of ORT to children's immunization cover in order to
explore the hypothesis that use of one form of effective health measure is likely to be part
of a package of effective measures. Again there is evidence of a strong association
between immunization and knowledge of effective treatment of diarrhoea in both
countries and for children of all ages. However, there is a dramatic difference between
the two countries in the proportions knowing of ORT. Almost all mothers in Zimbabwe
had heard of this treatment, compared with only 40 per cent in Burundi. This suggests
that whereas in Zimbabwe the EPI represents part of a comprehensive health campaign
12
which simultaneously addresses several aspects of child healthcare, the health
programme in Burundi has so far achieved results on a more limited front.
CONCLUSION
The above analysis depicts predictably higher levels of immunization coverage in
Zimbabwe, where there is better development of social infrastructure and more
economic resources are available to devote to healthcare promotion. Less
comprehensive health campaigns, along with civil unrest and lack of development of a
transport and communication infrastructure, have contributed to lower levels of child
immunization in Burundi. Although the data show little variation in immunization patterns
between educated and uneducated mothers, improvements in overall educational levels
in Burundi are likely to lead to improved timing of immunization and higher continuation
rates.
The most interesting feature of the above analysis is that, apart from generally higher
levels of immunization coverage among younger children born after the introduction of
the EPI, there is an absence of consistent differences between groups with different
characteristics. This must be considered a reflection of the successful promotion of the
EPI across different regions and different social groups in both countries, and suggests
that the attainment of complete immunization of all children now depends on service
availability rather than social acceptance.
13
REFERENCES
CALDWELL, John C.
1989
"Mass education as a determinant of mortality
decline"
Chapter 5 in John C. CALDWELL and Gigi SANTOW
(eds) Selected Readings in the Cultural Social
and Behavioural Determinants of Health
Health Transition Series No. 1
Canberra: Health Transition Centre,
The Australian National University pp101 -111
DEMOGRAPHIC and HEALTH SURVEYS
1987
Interviewer's Manual
Columbia: Institute for Resource Development /
Macro Systems
DEMOGRAPHIC and HEALTH SURVEYS and CENTRAL STATISTICS OFFICE
1989
Zimbabwe Family Health Survey I11988
Columbia: Institute for Resource Development /
Macro Systems
DEMOGRAPHIC and HEALTH SURVEYS and MINISTERE DE L'INTERIEUR
1988
Enquete Demographique et de Sante au Burundi,
1987
Columbia: Institute for Resource Development /
Westinghouse
HEGGENHOUGAN, Kris and John CLEMENTS
1987
"Acceptability of childhood immunization:
social science perspectives"
Evaluation and Planning Centre Publication No 14
London: London School of Hygiene and Tropical
Medicine
HENDERSON, R.H.
1984
"Vaccine preventable diseases of children:
the problem'1
in Protecting the World's Children: Vaccines and
Immunization within Primary Care
Conference Report, Bellagio, March, Rockefeller
Foundation pp 2-15.
MILLER,D.L., R. ALDERSLADE and I.M.ROSS
1982
"Whooping cough and whooping cough vaccines: the risks
and benefits debate"
Epidemiological Reviews, Vol. 4, pp 1-24.
MULLER
1984
"The epidemiology of pertussis and results of a vaccine
trial" in J.K. van Ginneken and A.S. Muller (eds)
Maternal and Child Health in Rural Kenya:An
Epidemiological Study",
London: Croom Helm, pp 95-108.
14
STOEKEL, Philippe
1984
"The Kolda-Kaya-Kolokani immunization programme"
in Protecting the World's Children: Vaccines and
Immunization within Primary Care
Conference Report, Bellagio, March, Rockefeller
Foundation pp 117-131.
STREATFIELD, P.K., Masri SINGARIMBUN and Ira SINGARIMBUN
"The impact of maternal education on the use of
1986
child immunization, and other health services".
Child Survival Research Note No 8CS,
International Population Dynamics Program,
Canberra: Australian National University.
UNICEF
1985
UNESCO
1988
Children and women in Zimbabwe: A Situation
Analysis
Government of Zimbabwe / UNICEF.
Country Program Recommendation: Burundi
United Nations Children's Fund Programme
Committee 1988 Session,
New York: UNESCO.
UNICEF
1990
The State of the World's Children, 1990
UNICEF: Oxford
ZIMBABWE NATIONAL FAMILY PLANNING COUNCIL
1985
Zimbabwe Reproductive Health Survey
Maryland: Zimbabwe National Family Planning
Council / Westinghouse Public Applied Systems.
Table 1: RECOMMENDED IMMUNIZATION SCHEDULES
FOR BURUNDI AND ZIMBABWE
BURUNDI
At birth
6 weeks
10 weeks
14 weeks
9 months
At some stage
BCG, Poll'd
DPT1, Polio2
DPT2, Polio3
DPT3, Polio4
Measles
DPT4
Source: EDSB, 1988: 70.
ZIMBABWE
At birth or soon after
3 months
4 months
5 months
9 months or soon after
BCG
DPT1, Poliol
DPT2, Polio2
DPT3, Polio3
Measles
Source: ZDHS, 1989: 86.
Note:
UNICEF (1985: 32) uses Dando's (1985) criterion
that a child is considered fully immunized if it received BCG at any time;
three doses each of DPT and polio at six weeks of age or later and
28 days apart; and measles at 8.5 or 9 months.
Table 2: PERCENTAGE OF CHILDREN AGED 12 - 23 MONTHS IMMUNIZED, ZIMBABWE, 1982 - 1988
1982
Chitungwiza
Bulawayo
Harare
1983
Outside cities
Zimbabwe
1984
1982
1984
1982
1985
1982
1984
1988
% in Age Group
With Health Card
83
88
90
92
85
91
71
81
82
% With Card
Fully Immunized
48
60
61
79
35
65
25
42
86
BCG record or scar
75
92
93
94
70
87
59
87
97
DPT1
DPT2
DPT3
70
66
57
84
80
77
85
76
65
94
90
80
76
67
48
88
84
83
57
45
32
76
74
66
98
96
92
Poliol
Polio2
Polio3
70
66
58
84
80
75
86
76
65
95
95
85
76
65
46
88
86
83
58
48
31
78
73
61
99
97
93
Measles
65
71
72
80
76
72
51
53
93
206
219
214
214
210
216
217
210
488
20
44
16
47
4
13
7
22
2
6
2
6
TYPE OF VACCINATION
n
DROP-OUT RATES
DPT2 / DPT1
DPT3 / DPT1
Polio2 / Poliol
Polio3 / Poliol
4
19
4
17
1
3
SOURCE: 1982-85 = UNICEF, 1985: 33 - 33;
1988 = ZDHS Datatape.
Table 3: PERCENTAGE WHO HAVE EVER
RECEIVED EACH VACCINATION
(all children)
BURUNDI
ZIMBABWE
38.5
36.8
31.4
26.5
7.3
37.9
32.2
27.0
16.3
25.3
62.9
59.0
56.2
52.3
TYPE OF
VACCINATION
BCG
DPT1
DPT2
DPT3
DPT4
Poliol
Polio2
Polio3
Polio4
Measles
n
3885
59.1
56.4
52.5
49.9
3576
Table 4:
PERCENTAGE OF E.P.I. RECOMMENDED VACCINATIONS RECEIVED
(all children, eight vaccinations)
ZIMBABWE
BURUNDI
%
%
CHILD'S
AGE
(months)
None
Up to 25
26 - 50
51
75
76 - 100
n
None
Up to 25
26 - 50
75
51
76 - 100
n
0 - 9
10 - 18
19 - 36
37 - 60
52.3
39.3
53.9
76.4
7.7
2.6
2.0
2.5
13.4
7.1
4.1
2.7
8.1
7.6
4.4
3.3
18.5
43.4
35.6
15.0
688
606
1153
1438
21.7
21.2
29.8
49.2
30.2
1.0
0.8
0.7
14.8
1.8
1.0
1.3
12.3
7.2
2.7
1.9
20.9
68.8
65.7
47.0
506
500
1024
1454
Total sample
59.6
3.3
8.0
6.2
22.9
3885
35.5
5.1
3.2
4.4
51.8
3484
Figure 1:
Proportions Immunized (With Card)
100
90
80
70
60
■ BURUNDI
50
■ ZIMBABWE
40
30
20
10
BCG
DPT1
Pol wl
DPT2
Polio2
DPT3
Polio3
Measles
BCG
MEASLES
250
250
200
200
\
150
150
100
100
50
50
0
0
2
4
3
5
6
7
8
rrrrmrTT^rT^,
o 4-
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
0
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
9
DPT1
DPT2
250
250
200
200
150
150
I
100
1
100
50 /
50
0
0
0
1
2
3
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
0
___ ^nTYTTTTTi rr i ■T~T^r—i
3
12
5 6 7
4
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
DPT3
DPT4
250
200
180
160
140
120
100
80
60
40
20
0 0 1
200
150
100
50
0 L
0
J
12
r~r~~r~r~i
3
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
■ T r~T—T-T—r-^TTI | | I
2
3
7
6
5
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
POLIO1
POLIO2
250
180
160
200
\
140
120
150
100
100
80
60
50
40
20
0
0
2
--^n~TTTTTTTXXa=^
0
3
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
0
POLIO3
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
POLIO4
200
180
160
140
120
100
80
60
40
20
250
200
150
100
50
0 10
2
3
5
6
-I ITTTte
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
FIGURE 2 : BURUNDI:
0
0
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
AGE AT VACCINATION
MEASLES
BCG
1400
600
1200
500
1000
400
800
300
600
200
400
100
200
IIIL i
0
0
0
1
2
3
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
0
2
3
4
5
6
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
DPT2
DPT1
900
800
800
700
700
600
600
500
500
400
J
400
300
300
200
200
100
100
0 J-
0
0
2
5
3
6
8
7
9
2
0
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
3
5
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
6
DPT3
POLIO1
800
900
700
800
600
700
600
500
L.
500
400
300
200
100
0
0
2
3
\
J
4
400
300
200
100
0
5
6
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
0
1
2
4
3
POLIO2
6
5
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
POLIO3
800
800
700
700
600
600
500
500
400
300
200
100
0 J0
1
J
2
3
4
400
300
200
100
0
5
6
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
FIGURE 2
0
2
ZIMBABWE:
3
4
5
6
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
AGE AT VACCINATION
TABLE 5: PERCENTAGE OF IMMUNIZATIONS RECEIVED AT RECOMMENDED AGE
(children with health card)
BURUNDI
ZIMBABWE
%
%
Chi Id's
age
(months)
None at
rec.
age
0 - 3
4 - 6
7 - 9
10 - 12
13 - 18
19 - 24
25 - 36
37 - 60
8.3
11.0
19.0
23.3
27.5
37.4
47.1
55.5
Up to 25 26 - 50
0.0
44.1
40.5
39.0
41.9
26.9
24.0
20.1
21.4
26.3
24.6
13.0
16.7
15.4
12.6
9.4
51
75 76
11.9
13.6
11.1
16.4
9.5
13.2
10.6
7.1
100
58.3
5.1
4.8
8.2
4.5
7.1
5.7
8.0
n
Child's
age
(months)
84
118
126
146
222
182
350
339
0 - 3
4 - 6
7 - 9
10
12
18
13
19 - 24
25 - 36
37 - 60
None at
rec.
age
0.0
3.8
5.8
3.1
4.2
3.0
7.2
13.3
Up to 25 26 - 50
0.0
16.9
18.3
26.0
20.8
25.1
21.1
27.4
29.9
16.1
15.0
11.0
13.5
19.5
13.9
18.6
51
75 76 - 100
n
68.6
21.5
42.5
40.9
45.9
38.1
41.2
29.9
137
130
120
127
259
231
469
715
1.5
41.5
18.3
18.9
15.4
14.3
16.6
10.8
Table 6: PERCENTAGE OF E.P.I. RECOMMENDED VACCINATIONS
RECEIVED BY MOTHER S EDUCATION
(all children, eight vaccinations)
BURUNDI
ZIMBABWE
CHILD'S
AGE
MOTHER'S EDUCATION
None Primary Higher
MOTHER'S EDUCATION
None Primary Higher
0 - 9 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
55.7
8.0
12.6
8.3
15.3
47.1
6.4
17.1
7.1
22.1
21.2
9.1
9.1
9.1
51.5
19.2
33.3
15.4
12.8
19.2
24.6
30.6
13.8
12.1
18.9
16.8
27.5
16.8
12.2
26.7
515
140
33
78
297
131
40.9
3.0
6.8
7.2
42.2
33.3
1.0
10.8
9.8
45.1
32.3
3.2
0.0
6.5
58.1
20.9
I. 2
II. 6
66.3
17.9
1.7
2.0
7.1
71.3
29.7
0.0
1.7
4.2
64.4
472
102
31
86
296
118
19-36 MONTHS
54.7
None
Up to 25% 1.7
26 to 50% 4.1
51 to 75%
4.8
76 to 100% 34.7
51.0
2.9
4.6
3.7
37.8
53.8
3.1
1.5
1.5
40.0
25.4
1.0
1.0
5.1
67.5
29.0
0.5
1.2
2.7
66.6
38.0
1.8
0.0
0.0
60.2
847
241
65
197
656
171
76.5
2.5
3.1
3.4
14.5
78.2
2.5
1.5
4.0
13.8
68.1
2.9
1.4
0.0
27.5
46.6
1.4
1.4
2.1
48.6
48.6
0.6
1.6
2.0
47.2
55.4
0.0
0.0
1.0
43.6
1093
275
69
290
960
204
TOTAL SAMPLE
None
60.8
Up to 25% 3.3
26 to 50% 5.7
51 to 75%
5.3
76 to 100% 25.0
57.8
3.2
6.6
5.3
27.2
50.0
4.0
2.5
3.0
40.4
33.5
4.9
2.9
5.5
53.1
35.4
4.8
3.2
4.3
52.4
37.7
6.3
3.8
3.7
48.6
758
198
651
2209
624
n
10-18 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
n
37 - 60 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
n
2927
Table 7: PERCENTAGE OF E.P.I. RECOMMENDED VACCINATIONS
RECEIVED BY PLACE OF RESIDENCE
(all children, eight vaccinations)
BURUNDI
CHILD'S
AGE
ZIMBABWE
URBAN
RURAL
URBAN
RURAL
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
42.9
7.6
10.5
9.5
29.5
54.0
7.7
13.9
7.9
16.5
12.3
33.1
12.3
12.3
30.0
25.0
29.3
15.7
12.2
17.8
n
105
583
130
376
39.0
1.3
6.5
5.2
48.1
39.3
2.8
7.2
7.9
42.7
21.5
0.8
2.3
75.4
21.1
1.4
2.2
8.9
66.5
77
529
130
370
61.9
1.3
1.9
1.9
33.1
52.6
2.1
4.4
4.8
36.1
33.6
0.8
0.8
64.8
28.6
0.8
1.3
3.4
66.0
160
993
250
774
79.3
1.4
0.9
0.9
17.4
75.9
2.7
3.0
3.8
14.6
49.6
0.5
0.5
0.7
48.7
49.0
0.8
1.6
2.3
46.3
213
1225
409
1045
61.8
2.5
3.8
3.4
28.5
59.3
3.4
6.0
5.5
25.8
36.0
5.1
2.1
2.6
54.2
35.3
5.0
3.7
5.0
51.0
555
3330
919
2565
0 - 9 MONTHS
10-18 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
19-36 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
37 - 60 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
TOTAL SAMPLE
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
Table 8: PROPORTION OF E.P.I. RECOMMENDED VACCINATIONS
RECEIVED BY HOUSEHOLD OWNERSHIP OF RADIO
(all children, eight vaccinations)
CHILD'S
AGE
0 - 9 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
10-18 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
19-36 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
37 - 60 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
TOTAL SAMPLE
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
BURUNDI
ZIMBABWE
OWNS RADIO
OWNS RADIO
Yes
No
Yes
No
33.8
5.6
16.9
8.8
35.0
58.0
8.3
12.3
8.0
13.4
18.5
29.9
13.3
11.8
26.5
24.1
30.5
15.9
12.5
16.9
160
528
211
295
35.9
2.3
6.1
6.1
49.6
40.2
2.7
7.4
8.0
41.7
22.7
0.9
1.4
4.5
70.5
20.0
1.1
2.1
9.3
67.5
131
475
220
280
49.8
1.5
1.8
3.7
43.2
55.1
2.2
4.8
4.7
33.3
28.7
0.3
1.6
69.5
30.4
1.2
1.4
3.4
63.5
273
880
383
641
74.1
1.7
2.6
2.0
19.6
77.2
2.7
2.7
3.8
13.6
47.0
0.5
0.9
0.9
50.7
50.6
0.8
1.6
2.5
44.6
347
1091
572
882
61.3
3.6
5.8
5.4
23.9
54.2
2.4
5.4
4.3
33.7
36.6
5.1
3.6
5.1
49.5
33.8
4.9
2.7
3.3
55.3
2974
911
2098
1386
Table 9: PERCENTAGE OF E.P.I. RECOMMENDED VACCINATIONS
RECEIVED BY TETANUS INJECTION BEFORE BIRTH
(all children, eight vaccinations)
ZIMBABWE
BURUNDI
CHILD’S
AGE
Yes
HAD INJECTION
No
Yes
HAD INJECTION
No
0 - 9 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
46.8
7.8
14.6
8.6
22.2
67.4
7.6
10.3
6.5
8.2
17.6
30.6
16.7
12.5
22.5
40.2
32.2
8.0
9.2
10.3
n
500
184
408
87
34.8
2.2
6.5
8.6
4s:o
48.6
3.9
8.3
5.5
33.7
18.6
0.7
1.9
6.7
72.1
33.3
2.9
1.4
10.1
52.2
417
181
420
69
52.3
1.2
3.4
4.5
38.6
56.1
3.4
5.4
4.1
31.0
26.7
1.0
1.0
2.0
69.4
40.2
731
410
817
184
72.8
2.3
2.8
3.2
18.9
77.2
3.3
2.9
4.0
12.6
38.4
0.7
1.4
2.2
57.2
53.3
1.1
1.9
1.5
42.1
750
578
982
261
54.5
3.1
6.1
5.7
30.7
65.6
4.0
5.4
4.6
20.4
28.4
5.4
3.7
4.5
58.0
45.1
5.5
2.5
5.2
41.8
2398
1353
2627
601
10-18 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
19-36 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
37 to 60 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
TOTAL SAMPLE
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
1.1
6.5
52.2
Table 10: PERCENTAGE OF VACCINATIONS RECEIVED
BY KNOWLEDGE OF ORAL REHYDRATION THERAPY
(all children, eight vaccinations)
CHILD'S
AGE
BURUNDI
ZIMBABWE
HEARD OF ORT
HEARD OF SUGAR-SALT
SOLUTION
No
Yes
Yes
No
None
Upto 25%
26 to 50%
51 to 75%
76 to 100%
39.5
8.8
13.8
10.3
27.6
60.0
7.1
13.2
6.8
12.9
19.4
30.8
15.2
12.7
21.9
52.9
29.4
11.8
5.9
0.0
n
261
425
480
17
30.0
1.6
7.4
7.0
54.1
45.7
3.5
6.9
8.1
35.8
19.4
0.8
1.9
7.2
70.7
57.1
14.3
0.0
14.3
14.3
257
346
485
7
49.9
2.7
4.5
3.9
39.0
56.0
1.6
3.8
4.8
33.8
28.4
0.8
1.0
2.7
67.0
33.3
0.0
0.0
5.6
61.1
441
707
988
18
75.0
1.9
2.4
2.8
17.9
77.0
2.9
2.9
3.8
13.3
46.3
0.7
1.4
2.0
49.6
71.4
0.0
0.0
0.0
28.6
577
848
1360
28
54.2
3.3
5.8
5.1
31.6
62.9
3.4
5.7
5.3
22.8
33.1
5.1
3.4
4.5
53.8
55.7
8.6
2.9
4.3
28.6
1536
2326
3313
70
0 - 9 MONTHS
10-18 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
19-36 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
37 - 60 MONTHS
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
TOTAL SAMPLE
None
Up to 25%
26 to 50%
51 to 75%
76 to 100%
n
Research Note Number
CHILD
SURVIVAL
Date
33CS
14 February 1991
Division of Demography and Sociology
Research School of Social Sciences
The Australian National University
Canberra ACT, Australia
A Project Sponsored by The Ford Foundation
CONTRACEPTIVE USE PATTERNS OF COMMUNITY FAMILY PLANNING
VOLUNTEERS IN INDONESIA
Dov Chernichovsky
The World Bank
Charles Lerman
and
Pudjo Rahardjo
The Indonesian National Family Planning Coordinating Board,
Jakarta
Note:
p/
'
\
-t’oiv
/
Child Survival Research Notes are brief discussions
of issues of current relevance to researchers and
policy-makers concerned with problems of high infant
and child mortality in the world. The International
Population Dynamics Program, Department of Demography,
The Australian National University, distributes these
notes with their regular Bibliographic Circular.
Production of the Child Survival Research Notes is
made possible through a grant from the Ford Foundation.
Responsibility for the content of Child Survival
Research Notes rests with the author(s) alone, and not
the above-listed organisations.
I)
J
. -■'VX
1990 - Celebrating Sixty Years of Higher Education in the ACT
ANU Open Day - Sunday 16 September 1990
«
4
Introduction
Community volunteers are often key actors in recruiting and
maintaining family planning acceptors at the grassroots level.
Studies have shown that highly motivated, knowledgeable, and
well-supervised community workers are in a prime position to
distribute non-clinical contraceptives, furnish basic family
planning information and advice, motivate potential acceptors to
adopt contraception, and refer clients who desire clinical or
surgical contraception or who suffer from side-effects to
appropriate health facilities (Ross, et al., 1987; Beeson, et
al., 1987; Kols and Wawar, 1982; Huber, et al., 1975; Foreit, et
al., 1978). They can be especially effective in introducing
young, low parity, poorly educated, and rural women to family
planning (Teachman and Rahardjo, 1979). Their participation can
thus be indispensable for the successful institutionalization of
national family planning programs.
A shift from public sector to commercial or privatized
family planning programs may signal an even greater contribution
by community-based distributors to program success. In
developing countries, barriers faced by villagers for using
clinic-based methods should continue in the foreseeable future.
These barriers include poor accessibility to clinics, high costs
of clinical methods, bureaucratic obstacles to receiving proper
instruction and care, and cultural and community objections to
clinic-based methods (Huber, et al., 1983). Assuring program
sustainability may thus necessitate maintaining workers at the
local level who handle family planning-related matters in an
accurate and timely fashion and at reasonable cost.
Over the past fifteen years, the Indonesian National Family
Planning Program (BKKBN) has promoted a community-based approach
to its family planning delivery system. According to the 1987
National Indonesia Contraceptive Prevalence Survey, almost
one-third of all pill/ condom, and injectable users obtain their
supplies from community-based sources (family planning field
workers, family planning posts, and integrated service posts).
Pills are the single most popular method in Indonesia, one-third
of total use, and 46 percent of pill users acquire their supplies
through community-based channels1.
Tracked historically during the 1970s, contraceptive
prevalence rose in tandem with a nation-wide expansion of the
community-based distribution system (BKKBN, 1989), and many
commentators have ascribed a close connection between the two
(Ross, et al., 1987; Sinquefield and Sungkono, 1979;
1
I
Chernichovsky and Meesook, 1981? Warwick, 1986). Family
planning volunteer workers or kader are key members of this
system at the local level.
In 1987, the Indonesian National Family Planning Program
started laying the groundwork of KB Mandiri, or the Family
Planning Self-Reliance Program. The Program’s objectives are
three-fold — first, to inspire a change of attitude among users
so that they make their own decisions about family
planning-related matters rather than have decisions made for them
by other parties; second, to boost participation in family
planning service delivery by private and commercial sectors and
also by the community? and third to provide services for those
who desire better quality care and who are willing to pay for
this care (Haryono Suyono, 1989). With some three-quarters of
the population still living in villages, achieving these goals
will involve integrating KB Mandiri activities into existing
community-based distribution networks. Under this new system,
however, local providers should become increasingly instrumental
for sustaining program success. For purposes of recruiting new
village providers and maintaing current ones, it is thus
imperative for the program to understand community provider
characteristics and the ways in which these characteristics
relate to attaining desired program goals.
This paper explores the characteristics of Indonesian
family planning volunteer workers. It covers three areas: (1) a
description of community family planning volunteer
characteristics, (2) an assessment of the impact these
characteristics have on the likelihood of volunteer contraceptive
use and method choice, and (3) an evaluation of the impact field
worker contraceptive use and choice have on volunteer use and
choice. An evaluation is also made on the likely effect that
both environmental and program policy and design factors have on
volunteer contraceptive use and choice. Finally, an assessment
is made as to how this information can be useful to program
managers charged with furthering family planning self-reliance at
the local level.
Study Areas and Coverage
A field survey was conducted in 1986/1987 in order to obtain
information on the cost and effectiveness of the national family
planning delivery system. The study area included six regencies
in three provinces: Tangerang in West Java, Kulon Progo and
Bantul in the Special District of Yogyakarta (Daerah Istimewa or
DI Yogyakarta), and Banjar, Barito Kuala, and Tapin in South
Kalimantan. Because there are large interregional variations in
both socioeconomic conditions and family planning program
designs, results from these six regencies should not be taken as
representative of other areas in Indonesia.
2
9
Even though the survey gathered information from the entire
population of field-level family planning health, program, and
community personnel, this study only uses data provided by field
workers and volunteers. Field workers furnished information
about both communities and volunteers in their catchment areas.
One should note that the management structure of the
National Family Planning Program encourages decentralized
planning and decision-making (Haryono Suyono and Shutt, 1989).
When the community-based distribution system was initially
planned over fifteen years ago, program managers made a conscious
decision not to standardize it, arguing that regional cultural
and social variations should dictate different program designs
and implementations.
This decentralization makes for a highly complex program.
All of Indonesia’s provinces are subdivided into regencies and
municipalities which are further partitioned into subdistricts.
BKKBN assigns only one family planning field worker supervisor
per subdistrict. The next administrative subdivision is the
village. On average, BKKBN allocates one field worker to cover
four to six villages, although this allocation can vary according
to the size of the eligible couple coverage area and geographical
terrain. Typically, there are several volunteer workers per
village, one or more of whom could work in the family planning
arena. Depending on demographic and program characteristics of
given areas, there can be a further administrative subdivision
into hamlets or even neighborhoods that are served by subvillage
volunteer workers.
Community family planning volunteers operate in units known
as Subdistrict Management Units or PPKBD fos Pembina Keluarga
Berencana Desa); at the subcommunity level, these units are known
as sub-PPKBD.
(Subdistrict Management Units have erroneously
been translated as Village Contraceptive Distribution Centers or
VCDC in English.) Personnel in these units assist village or
hamlet leaders in family planning-related matters. Supervised by
family planning field workers, their membership typically
includes representatives from the Women’s Welfare Group (PKK),
acceptor groups, and community volunteers (Haryono Suyono and
Shutt, 1989).
Table 1 shows that considerable regional variation may exist
in the distribution of community and subcommunity personnel.
In
DI Yogyakarta, subcommunity volunteers constitute 92 percent of
the total, whereas the comparable figure in South Kalimantan is
only 44 percent. These differences may occur for three reasons:
different complexities of administrative structure by province;
different program maturities by region; and different types of
village organization by area.
3
Village leaders (lurah) typically recruit the family
planning volunteers, but BKKBN guidelines govern their choices.
These guidelines stipulate that volunteers should have a minimum
of a primary school education, be prominent or exemplary members
of the community, and use contraception themselves. Either men
or women may be selected as volunteers, although in practice
women are usually preferred. Since BKKBN places heavy emphasis
on the leadership capabilities of these volunteers, they are
commonly the wives, relatives, or social clients (anak buah) of
the village leaders. The Government of Indonesia has generally
followed a policy of not paying kader, claiming that to do so
could undermine their ascriptive social role of promoting
community development2.
Volunteer duties include distributing pills and condoms to
acceptors, motivating potential new acceptors, and making
referrals to clinics.. In addition, volunteers should record
contraceptive use on village maps and make monthly reports to
field workers, oversee village contraceptive stocks, recommend
IUD use to pill and condom users, disseminate correct information
about contraceptives, and assist family planning acceptor group
members and personnel in the integrated health posts (Judd,
1987A; Judd, 1987B; BKKBN, 1977).
Community Family Planning Worker Characteristics
Table 2 shows kader social and demographic characteristics
by study area.
Few age differences distinguish community volunteers in the
various areas, although those from DI Yogyakarta are slightly
older on average than their counterparts in the other provinces.
Bantul volunteers have noticeably higher education than those
from the other areas, especially Tangerang volunteers. The
proportion of volunteers who are female is comparatively high in
DI Yogyakarta and comparatively low in South Kalimantan.
Somewhat higher proportions of family planning volunteers use
contraception in South Kalimantan compared with Tangerang and DI
Yogyakarta.
IUDs are the method of choice in DI Yogyakarta,
followed by female sterilization. Tangerang volunteers prefer
injectables to the next most prevalent method, pills. Most South
Kalimantan volunteers use pills, with the next most common method
injectables.
Interpretation of the Descriptive Results
Data from the 1980 Indonesian Population Census provide
evidence about characteristics of the population from which
community workers are selected (not shown). Tangerang and the
South Kalimantan regencies have a young age structure compared
with the DI Yogyakarta regencies, and Tangerang and Banjar, in
particular, have male-dominant sex ratios. This male dominance
4
is most likely the result of sex-selective migration
patterns. Tangerang has a notably higher illiteracy rate than
other regencies and also higher proportions of its population
over the age of 10 with only elementary school education or less.
These differential characteristics also appear among community
family planning volunteers. Kader attributes thus seem to
reflect many of the attributes associated with different
population bases.
Table 3 shows contraceptive use and method mix rates for the
six regencies according to BKKBN*s service statistics. The
service statistics cover women of average younger age than the
volunteers which may account for the large discrepancy between
the two data sources for the percent not using contraception or
using the less permanent methods.
Both the survey and the service statistics show high
injectable use in Tangerang, high IUD use in DI Yogyakarta, and
high pill use in South Kalimantan., <Community volunteers, within
<certain demographic and social parameters, thus appear to mirror
in their own contraceptive choices the contraceptive choices and
tastes of their clients.
Areal influence, however, is only one factor in
understanding distributions of volunteer characteristics,, As
mentioned above, 1BKKBN
---- ..has its own set of recommendations for
volunteer recruitment, Because volunteer characteristics lack
uniformity across areas, it is clear that the application of
these recommendations is tempered by local environmental and
labor market conditions, but they nevertheless contribute to the
selection process.
Correlates of Contraceptive Use and Method Choice
The client-provider interaction is complex and involves many
cultural, psychological, and programmatic factors. In attempting
to unravel it, there is an assumption that particular measurable
traits and activities of family planning workers are associated
with better promotion of family planning.
In the Indonesian context, "better promotion of family
planning" entails (1) motivating eligible couples to use
contraception and (2) motivating eligible couples to use
preferred contraceptive methods. The most cost-effective
methods^ (US$ cost/CYP) are IUDs and sterilization (Lerman,
1987), but of the two only IUDs are currently included in the
public sector National Family Planning Program. BKKBN strongly
encourages preferred method use among its field workers, field
worker supervisors, and community family planning volunteers.
5
f
It is assumed that kader who use contraception themselves
are more likely to recruit and maintain clients than those who do
not use contraception. It is also assumed that kader who use
preferred program methods are more likely to recruit and maintain
clients who use these methods. Because of data constraints,
there is no direct test of these assumptions in this study, but
empirical studies exploring "homophilous1’ effects both in
Indonesia and outside the country strongly suggest that field
workers who have characteristics in common with their clients are
more effective than those not sharing these characteristics
(Repetto, 1977; Beeson, et al., 1987).
Two reasons could explain this matched fit of
characteristics between kader and clients. First, the high
credibility afforded to community family planning volunteers on
the basis of their social standing may induce others to follow
their lead. This demonstration effect is likely to be strong in
Indonesia where villages tend to be tight-knit and where
community leaders spearhead family planning acceptance through
their face-to-face encounters with potential clients. As members
of the same community, these leaders would likely be more
effective in legitimizing family planning than outsiders.
Second, as shown by comparing survey data in Table 1 and
service statistics data in Table 2, contraceptive method mix
rates for volunteers and clients by region are generally alike.
To the degree that both volunteers and clients live in the same
social, economic, and cultural environment and to the degree that
they are exposed to the same program service and delivery
systems, their contraceptive use and choice patterns should in
fact be similar.
Because family planning practices vary so greatly across
regions, the analysis is performed holding regions constant so
that interregional differences along do not explain total
variance. Two of the most important variables thereby controlled
are differential contraceptive availability and differential
program design which by themselves could largely explain kader
contraceptive choice.
The analysis is divided into two parts:
A)
A model in which the dependent variable is binomial; whether
or not a family planning volunteer currently uses family
planning (1 = use; 0 = non-use).
B)
A model in which the dependent variable is multinomial; the
contraceptive method used by a family planning volunteer (0
= IUD; 1 = pill; 2 = condom; etc.)
6
The first model employs a binomial probit model (Theil,
1971). Its objective is to determine conditional probabilities,
in this case to determine whether or not a volunteer practices
family planning conditioned upon the following explanatory
variables: age, number of living children, level of education,
and whether or not the volunteer’s field worker practices family
planning. Dummy variables are also included to control for the
effect of regency variation within the provinces of DI Yogyakarta
and South Kalimantan.
The second model employs multinomial logit analysis which
permits predicting the odds of preferring a given method compared
with a reference method (Maddala, 1983). In this study, the
reference method is the IUD. The explanatory variables are the
same as in the probit analysis except for the field worker
variable, which in this model is whether or not volunteers use
the same method as their field workers.
Findings on the Determinants of Contraceptive Use
The estimated coefficients in the probit analysis reveal
that the relationship between age and the odds of using
contraception is complex, exhibiting significance in both linear
and curvilinear forms (Table 4). The linear relationship shows
that as volunteers grow older, they are more likely to use
contraception. This finding indicates that increasing age is
associated with a desire to terminate childbearing. The
curvilinear relationship suggests that, all things being equal,
the odds of using contraception rise up to the age of 20 in West
Java, 25 in DI Yogyakarta, and 23 in South Kalimantan, and level
off or decrease thereafter. Decreasing odds of using
contraception as a function of age could occur for several
reasons: declining use of contraception with approaching
menopause, increasing subfecundity among older women, increasing
divorce and widowhood among older women, and a cohort effect of
older women being less likely ever to have used contraception.
The odds of using family planning rise with parity which may
suggest that family planning use is associated with reaching
desired numbers of children.
The relationship between education and the odds of using
contraception is insignificant. As program volunteers,kader
would always have contraceptives available to them, no matter
what their educational level. Furthermore, as a general rule,
education probably has the greatest influence when family
planning is regarded as a matter of private discretion, rather
than as a collective responsibility, and when government
intervention is narrowly defined. In Indonesia, however,
community-based distribution networks, strong government support,
and program targets ensure that family planning supplies and
7
activities are funneled to all segments of the population,
regardless of educational level. This type of system clearly
tends to dilute the impact of education.
The data support the notion that field worker use of
contraception has a positive impact on kader use. This
relationship is significant in both West Java and South
Kalimantan. There are no significant differences in
contraceptive use between the regencies within DI Yogyakarta or
South Kalimantan.
Findings on the Determinants of Method Choice
The multinomial logit analyses in Tables 5 to 7 show the
relationship of the explanatory variables with the odds of
choosing a given contraceptive method compared with the IUD.
Older kader are generally more likely to use sterilization than
IUDs and younger kader pills than IUDs, but these relationships
are not statistically significant. In West Java, the relative
probability of using injectables vs. IUDs decreases with
increasing age up to age 37, which suggests that in this regency
volunteers gravitate toward more effective contraception with
age. In all three regions, numbers of living children are
negatively correlated with the odds of choosing pills over IUDs.
High parity, however, is positively associated with the choice of
female sterilization over IUDs in West Java and DI Yogyakarta.
These results suggest that volunteers rationally choose the
most permanent program methods as they grow older and after they
decide to terminate their childbearing; this choice, however, is
likely to be strongly influenced by the program objective of
promoting preferred contraceptive use among its own personnel.
Relationships between education and the odds of choosing a
particular contraceptive method are generally weak. In all three
areas, family planning volunteers with junior high school
education are less likely than those with no schooling to choose
pills over IUDs. In West Java, those with primary and senior
high school education are also less likely to choose pills over
IUDs. Volunteers with junior and senior high school educations
are less likely to choose injectables over IUDs in DI Yogyakarta.
Since IUDs are the favored method of choice in DI Yogyakarta and
injectables the favored method in West Java, these findings
suggest that those who choose against the prevailing trend
possess somewhat higher educational levels.
The results of the effect of field worker's method choice
are mixed and inconclusive. In Table 5, the negative coefficient
in the sterilization columns indicates that in DI Yogyakarta if
field workers are sterilized there is a significantly less
relative chance of volunteers being sterilized than if field
workers use an IUD and volunteers also use an IUD. The large
8
6
majority of DI Yogyakarta field workers and community
workers use IUDs, which implies that for field workers who choose
the IUD, there is a probability of volunteers making the same
choice. In West Java, even though most field workers use the
IUD, the majority of community workers use injectables, but here
also, field worker use of IUDs appears to have a positive impact
on volunteer use. It is difficult, however, to infer causality
from these particular findings which essentially reflect the low
proportion of field workers in these provinces who do not use the
IUD . For both DI Yogyakarta and West Java, field worker use of
methods other than IUDs has little positive influence on the use
of these methods by volunteers.
In South Kalimantan, field worker use of pills increases the
likelihood of volunteer use of pills in contrast to field worker
use of IUDs increasing the likelihood of volunteer use of IUDs.
Field worker use of sterilization also has a positive impact on
volunteer use of sterilization.
In this province, relatively few
field workers and volunteers use IUDs.
Some differences in volunteer method use appear between the
regencies in DI Yogyakarta and South Kalimantan. Volunteers in
Kulon Progo are less likely to use pills and injectables than
volunteers in Bantul, and volunteers in Tapin are less likely to
use pills than Banjar volunteers. These findings may indicate
variations in policy direction and implementation between regency
programs.
Conclusions and Policy Recommendations
The interpretation of these findings indicates a complex
interweave of environmental, programmatic, and individual
influences in determining the recruitment of volunteers and the
adoption by these volunteers of specific contraceptives. First,
community family planning volunteers are drawn from local labor
markets and thus possess many of the personal attributes of the
base population. Their choice of contraceptives may thus reflect
the prevailing choices of eligible couples in their coverage
areas. At the same time, BKKBN has developed a set of
demographic, social, and contraceptive use recommendations for
volunteer recruitment. Finally, there is an element of
self-selection by individuals who become volunteers, and these
characteristics have a bearing on contraceptive choice.
The specific contribution of each influence is difficult to
determine and in any case probably varies by area. The
complexity of the situation is highlighted by the issue of
volunteer contraceptive choice. The rough correspondence between
eligible couple and volunteer method use suggests that ambient
characteristics affect volunteer choice. This environmental
influence also comes to light in areal differences in volunteer
method choice between provinces and between regencies within
9
provinces. Volunteers, however, are also actors, and, given
their leadership positions, they undoubtedly help determine or
guide client choice. What is clear in the Indonesian context is
that individual choice cannot be entirely divorced from community
choice.
The relationship between family planning field workers and
volunteers appears to have two components: a demonstration effect
and a program policy effect. Volunteers are more likely to use
contraception if their field workers use contraception; less
conclusive evidence suggests that the same holds true with
specific method choice. This correlation may be partly explained
by a demonstration effect, with volunteers being influenced by
the personal example and prestige of their field workers. A
related explanation is that field workers communicate program
policies about preferred methods to volunteers, actively
motivating them to use these methods.
(One indication that the
field workers deem this policy important would be their own use
of preferred methods.) Yet another explanation is that both
field workers and volunteers are recruited on the basis of their
using these methods.
Finally, the multivariate analysis indicates that volunteer
characteristics, such as age, parity, and education, are
correlated with method use and choice. Volunteers thus appear to
exercise latitude of choice independent of choices dictated by
program policies and design. Analysis of the 1987 National
Indonesia Contraceptive Prevalence Survey shows that age and
education are also important correlates of acceptor contraceptive
use and choice (Molyneaux et al., 1989), which is an indication
that volunteers and their clients may make contraceptive choices
for similar reasons.
The National Family Planning Coordinating Board is now
publicly committed to instituting a family planning
self-sufficiency, or KB Mandiri, program. Transferring a large
share of family planning from the public to the private arena,
this shift will entail greater participation by private
physicians and midwives, private contraceptive suppliers and
distributors, and community leaders in family planning efforts.
It will also accent the social marketing of non-clinical
contraceptive products. In launching this program, BKKBN assumes
that Indonesians have now embraced a small family norm, and that
they can exercise latitude in choosing suitable contraceptive
methods for themselves. This particularly holds true for those
who can afford family planning; for those who cannot, BKKBN and
local communities will still partially or fully bear the costs.
What are the specific implications of these findings for the
institutionalization of the KB Mandiri program? First, BKKBN
administrators should remain cognizant that any recruitment of
paid community distributors will occur in local labor markets and
10
J
that policies will need to be tailored to local conditions.
This would argue against standardized hiring procedures applied
to the entire country. Second, BKKBN should maintain its policy
of encouraging preferred contraceptive methods for its field
workers. In addition, it should make reasonable efforts to
persuade community distributors to use preferred methods, while
giving due respect to their religious or cultural sensitivities
about IUDs or sterilization.
Third, since older, higher parity volunteers are both more
likely to use contraception and to use preferred methods, special
priority should be given to recruiting these individuals as
community distributors. This personal experience may increase
the likelihood that volunteers will refer clients for
clinic-based methods, despite their income coming from the sales
of non-clinical methods such as pills and condoms. Fourth, more
educated volunteers may gravitate toward use of the preferred
methods and they also may be more discriminating in their choice
of contraceptives. This variable, however, bears greater
scrutiny. Education, as an influence on contraceptive choice,
should become stronger as the program evolves toward greater
privatization and commercialization.
This study, like many before it, highlights the rich
diversity of Indonesia’s family planning program. To be
successful, the program must continue to respond to a mosaic of
local conditions and needs. KB Mandiri, by injecting more
choice, flexibility, and control into the current system, should
enhance the notable achievements already attained by the program.
11
FOOTNOTES
*
This paper is based on an operational research project
financed by the Dutch Government and conducted by the World
Bank and the Indonesian National Family Planning
Coordinating Board. The views in this paper are those of
the authors and not of the institutions involved. The
authors would like to thank Elizabeth Frankenberg, Andrew
Kantner, Marc Mitchell, John W. Molyneaux, and Haryono
Suyono for their helpful comments.
1
These should be taken as minimum percentages. Seventeen
percent of all pill users claim that they receive their
supplies through ’’other" sources. Even though the NICPS
authors speculate that this category may include friends and
relatives, the ’’other’’ category probably contains some
fraction of users who received pills through community
providers.
(See National Indonesia Contraceptive Prevalence
Survey 1987, Tables 4.1 and 4.5.)
2
Volunteers charge a nominal fee for their services in some
areas of the country, but this money is set aside for
community development.
3
The calculation covers only commodity costs of these
methods.
4
These particular results are strongly influenced by the
structure of the data and the use of dummy variables. 1There
are several volunteers associated with any given field
worker with a resultant, but unknown, loss of degrees of
freedom. Because there are relatively few field workers who
use methods other than the IUD, there is a low probability
for any group of volunteers not using the IUD to have a
field worker who behaves in the same way.
12
4
REFERENCES
Badan Koordinasi Keluarga Berencana Nasional (BKKBN). 1989.
BKKBN Annual Feedback Recapitulation Reports, 1971/72 1988/89.
Badan Koordinasi Keluarga Berencana Nasional (BKKBN).
Buku Baku PPKBD.
1977.
Beeson, Diane, M. Faith Mitchell, Helene L. Lipton, Donald H.
Minkler, and Philip R. Lee. 1987.
"Client-Provider
Transactions in Community-Based Family Planning Programs and
the Outreach Component of Clinic-Based Systems." Robert J.
Lapham and George B. Simmons (eds.), Organizing for
Effective Family Planning Programs. Washington, D.C.:
National Academy Press, 457-484.
Chernichovsky, Dov and Oey A. Meesook. 1981. Regional Aspects
of Family Planning and Fertility Behavior in Indonesia.
World Bank Staff Working Paper No. 462. Washington, DC: The
World Bank.
Foreit, James R., Martin E. Gorosh, Duff G. Gillespie, and C.
1978. Community-Based and Commercial
Gary Merritt.
Contraceptiive Distribution: An Inventory and Appraisal."
Population Reports, Series J, No. 19. Baltimore, MD: The
Johns Hopkins University Press.
Haryono Suyono.
1989. BKKBN and the Expanding Role of Private
Sector Family Planning Services and Commercial Contraceptive
Sales in Indonesia, Jakarta: National Family Planning
Coordinating Board.
Haryono Suyono and M.M. Shutt. 1988. "Strategic Planning and
Management of Population Programs: An Indonesian Case
Study." Paper prepared for the Eleventh International
Conference of the International Council on Management of
Population Programmes, Beijing.
Huber, Sallie Craig, P.T. Piotrow, Malcom Potts, Stephen L.
Isaacs, and R.T. Ravenholt.
1975. "Contraceptive
Distribution - Taking Supplies to Villages and Households. ii
Population Reports, Series J, No. 5. Baltimore, MD: The
Johns Hopkins University Press.
Judd, Mary. 1987A. Village Kader Study: An Investigation of
Kaders in Five West Java Villages, prepared for
USAID/Jakarta.
Judd, Mary. 1987B. Kaders in Indonesia: A Literature Review,
prepared for USAID/Jakarta.
13
(
Kols, Adrienne J. and Maria J. Wawar.
1982. "Community-Based
Health and Family Planning." Population Reports, Series L,
No. 3. Baltimore, MD: The Johns Hopkins University Press.
Lerman, Charles. 1987. "Cost Effectiveness Analysis of Program
Contraceptive Methods," prepared for USAID/Jakarta.
Maddala, G.S. 1983. Limited Dependent and Qualitative Variables
in Econometrics. Cambridge: Cambridge University Press.
Molyneaux, John W., Charles Lerman, E. Srihartati P. Pandi, and
Soni Trisno Wibisono. 1989. "Correlates and Determinants of
Contraceptive Method Choice in Indonesia." Paper presented
at the Population Association of America Meetings,
Baltimore, MD.
National Indonesia Contraceptive Prevalence Survey 1987.
Columbia, MD: Institute for Resource
Development/Westinghouse.
1989.
Repetto, Robert. 1977. "Correlates of Field Worker Performance
in the Indonesian Family Planning Program: A Test of the
Homophily-Heterophily Hypothesis." Studies in Family
Planning, 8,20:19-21.
Ross, John, Donald J. Lauro, Joe D. Wray, and Allan G.
Rosenfield. 1987.
"Community-Based Distibution." In
Robert J. Lapham and George B. Simmons (eds.), Organizing
for Effective Family Planning Programs. Washington, D.C.:
National Academy Press, 343-366.
1979. "Fertility
Sinquefield, Jeanne C. and Bambang Sungkono.
and Family Planning Trends in Java and Bali. ii International
Family Planning Perspectives, 5,2: 43-58.
Teachman, Jay D. and Pudjo Rahardjo. 1979. "Contraceptive Use
in the Indonesian Village Distribution System: Continuation
and Effectiveness." International Family Planning
Perspectives, 5,2: 66-72.
Theil, H.
1971. Principles of Econometrics.
Wiley & Sons.
New York: John
Warwick, Donald P.
1986. "The Indonesian Family Planning
Program: Government Influence and Client Choice."
Population and Development Review, 12,3: 453-490.
14
J
I
Table 1. Jianoer and Percent (in Parentheses) of Cowunity Fanily Planning Volunteers IPPX60) and Subcomunity Fanily
.Planning Volunteers (Sub-PPK80) by Region.
west Java
01 yogyaxaru
South hliuantan
CcwDity Fully Planning Volunteers
9.124
(18.4)
550
(7.7)
2.270
(55.8)
SubcciMaity Fanly Planning Volunteers
40.555
(81.6)
6,567
(92.3)
(44.2)
49.679
(100.0)
7,117
(100.0)
4,070
(100.0)
Totai
1.800
Source: SF.K8K Monthly Fieldvorker Reports, October 1987.
Table 2. Social -oa Ceiograohic Characteristics, Contraceptive Use, and Method Choice of Coeiunity Feiily Planning Volunteers
by Regency.
V. Java
Tangerang
01 Yogyakarta
Ku ion Progo
36.8
29.1
90.6
3.4
11.3
42.1
18.8
93.6
3.7
21.0
Percent Hot Jsinq
Percent IUD
Percent Pill
Percent Comob
Percent injectaole
Percent Hale Sterilization
Tercent Feiale Sterilization
Percent Other*
19.3
21.6
10.3
0.3
42.4
2.0
4.1
0.0
23.2
56.7
6.5
1.9
2.1
0.8
8.1
0.7
Total Percent
100.0
N
1303
01 Yogyakarta S. Kaliiantan
Bantu 1
Banjar
S. Kalinantan
Santo Kuala
S. Kalimantan
Taoin
TOTAL
Socitl ut
ionic
Characteristies
Mean Age
Percent Males
Percent Married
Mean Kunoer :f Chiioren
Percent Hign Scnool
39.4
32.G
87.3
24.2
36.1
51.7
85.7
2.8
17.9
18.9
• 48.7
9.9
3.1
10.4
1.1
7.0
0.9
15.0
5.1
69.5
1.6
5.9
0.0
1.9
1.0
14.7
6.1
68.7
1.1
6.9
0.8
1.0
0.7
10.8
10.9
68.5
1.2
5.9
0.0
0.7
2.0
19.3
29.6
23.3
1.2
20.1
1.9
4.9
0.6
100.0
100.0
100.0
100.0
100.0
100.0
658
689
362
271
196
3482
39.9
20.2
94.5
3.2
34.7
39.2
54.8
90.1
3.4
2.3
21.3
39.1
30.3
90.;
3.3
21.2
and Above
Cwtriceotire Use ind
KttM Choice
• Induces 'loiants ano traditional Mthocs.
I
<
Table 3. ’KKBl Service Statistics Contraceotive Use ano Method Mix Rates by Regency. January 1187.
I. Java
01 Yogyakarta
Tangerang Kulon Progo
01 Yogyakarta
Bantu 1
S. Kaiiiantan
Banjar
S. Kaliiantan
Barito Kuala
S. Kaliiantan
Taoin
32.2
39.9
0.9
54.6
0.8
2.7
1.1
Contrauotive t'Sc
Method Mix Rates
Percent lot Using
Percent 100
Percent Pill
Percent Condos
Percent Injectacie
Percent Other*
39.1
4.7
15.4
0.3
40.3
28.8
31.5
38.0
8.9
18.2
1.9
24.6
1.2
Total Percent
Nusoer of Eligwie
Couoles
42 J
2.2
49.1
4.2
22.1
• 15.8
3.1
2.9
0.1
1.7
59.5
0.8
4.8
'1.0
100.0
100.0
100.0
100.0
100.0
100.0
287,029
$6,188
100.282
62.126
30,377
19.372
1.2
4.7
• Includes Hie sterilization, feiale sterilization, vaginal tablets, and ieolants.
t
Iu;e 4. Frobit Regression Coefficients on the Odds of Comnity Fully Planning Volunteer Use of Contraceotion
(ismotic'^statistics in oarentheses).
Ust Java
(N:1074)
DI Yogyakarta
1 11:766 )
South Kahianun
Ace
0.039
(4.370)
0.086
(6.955)
0.074
(5.647)
Age imres
-0.001
(-5J69)
-0.002
(-7.901)
-0.002
(-6.506)
0.984
(7.160)
0.122
(3.909)
0.153
(3.832)
rfiiirv School
0.157
(1.386)
0.018
(0.100)
0.287
(1.511)
Jr. Hign School
-0.034
(-0.242)
0.135
(0.707)
0.391
(1.287)
Sr. Hica School
0.185
(0.138)
0.370
(1.875)
0.191
(0.866)
Acideiw University
-0.038
(-0.019)
0.163
(0.487)
-0.088
(-0.178)
0.260
(2.610)
0.115
(1.013)
0.278
(1.981)
luicer of Living Children
(N:376)
EducHion*
Field forrer Using Contraceotioa**
-0.066
(-0.644)
sutnl
Birito (sill
-0.057
(-0.373)
Twin
0.122
(0.537)
Log-. 1 Uli rood
it
nt
-460.830
-HLZSO
Omv Variable: ExclMtd category is no schooling.
Ottiiv Variable: Exclnded category is field worker not esing contnceotion.
Oimv Variable: Eadided category for 01 Yogyakarta is Kelon Progo Regency.
Excleded category for South Kaliiantan is Sinjar Regency.
-218.710
I
e
'■ole
Rultinciiil logit degression Coefficients on the Odds of Conenity Faiily Planning Volunteers Using a
Articular Hethod Coooared With the IUD. Tangerang Regency. West Java (asvitotic t-statistics in oarentheses).
Pill
Condoi
Injectable
Rale Sterilization
Fewie Sterilization
-0.1H
(-1.263)
-0.224
(-0.451)
-0.268
0.402
(-2.662)
0.315
(0.941)
(1.402)
0.003
(1.512)
0.003
(0.524)
0.004
(2.683)
-0.003
(-0.658)
-0.005
(-1.241)
-0.163
(-1.988)
-0.298
(-0.811)
-0.025
(-0.454)
0.112
(1.562)
0.319
(3.312)
Pnury School
-1.121
(-3.105)
-0.451
(-0.002)
-0.236
(-1.011)
-0.084
(-0.168)
0.132
(0.282)
Jr. High School
-1.112
(-2.982)
10.180
(0.066)
-0.839
(-3.103)
-1.831
(-1.684)
0.381
(0.118)
Sr. High School
-2.119
(-4.110)
11.133
(0.068)
-1.014
(-3.640)
0.123
(0.192)
0.654
(1.120)
Acaoeiy/Umversity
-23.192
(-0.001)
11.798
(0.012)
-22.995
(-0.001)
-20.779
(-0.001)
-20.579
(-0.001)
Field Worker Using
Contraceotion*'
-3.532
(-5.838)
-11.391
(-0.083)
-1.242
(-6.856)
-1.835
(-3.215)
-14.123
(-Q.C62)
Constant
3.908
(1.547)
-8.114
(-0.054)
6.321
(3.514)
-10.164
(-1.511)
-11.143
(-2.013)
Sccared
tawr of Liiieg Children
1
rcncatiotf*
I
!
i
log likelihood=-943.500: M:1074.
»«
i
Gueay Variable: Excluded category is no schooling.
Ouny Variable: Exclided category is field worker not using contraceotion.
t
4
hi:!« •ultfiiMHl Logit Regression Coefficients on the Mds of Comnity fully Phrnimg volunteers Using a Particuiar
wetrea Cocared nth the IUD. Bantul and Kulon Progo Regencies, 01 Yogyakarta (asyitotic t-statistics in oarentnesesi.
Pill
Ccnaoi
Injectable
Male Sterilization
Feule Sterilization
-0.094
(-0.549)
-0.139
1-0.513)
0.280
(1.139)
0.773
(1.184)
0.283
(1.137)
0.001
(0.554)
0.002
(0.722)
-0.004
(-1.327)
-0.001
(-1.203)
-0.004
(-1.382)
-0.247
(-2.460)
-0.185
(-1.234)
0.155
(1.467)
0.166
(0.833)
0.330
(4.125)
-0.616
(-1.425)
-1.009
(-1.286)
1.732
(1.615)
-0.090
(-0.078)
0.073
(0.148)
-1.020
(-2.226)
-8.070
(-0.096)
1.733
(1.610)
0.217
(0.190)
0.166
(0.330)
Sr. Higa School
-0J67
(-1J03)
0.026
(-0.035)
2.136
(2.003)
-0.048
(-0.040)
0.264
(0.525)
Aoaeiy/University
-1.173
(-1.248)
0.963
(0.955)
1.765
(1.303)
-10.598
(-0.025)
-8.415
(-0.153)
-17.576
(-0.035)
-15.921
(-0.040)
-17.646
(-0.031)
-2.888
(-2.754)
-2.381
(-7.202)
-0.886
(-3.395)
-0.565
(-1.378)
-1.965
(-5.477)
0.499
(0.801)
(-0.895)
2.796
(0.852)
0.637
(0.120)
7.144
(1.503)
-18.796
(-1.145)
-7.635
(-1.178)
Age Saaarec
Uoatar of Living Children
Education*
Prian Scnool
Jr. Hies School
Field worker Using
Ccotracecxion’*
Regercr4"
Klien Prego
Constint
Log Liielih«»:-774.52: 11=766.
®
2umv nrinble: Excluded category is no schooling.
Zunev finable: Excluded category is field worker not using contraceotion.
■•any Variable: Exc tided category is Banti I Regency.
-0.226
Tide 1. Muitifloiial Logit Regression Coefficients on the Odds of Commtv Faiily Planning Volunteers Using a Particular
Ketr.cd Ccioared With the IUD. Banjar. Kuala Santo, ana Taoin Regencies, South KaliBantan (asyitotic t-statistics in
csrentnesesi.
Pill
Injectable
N&le/Fei&le Sterilization
Ace
-0.030
(-0.128)
0.155
(1.008)
-0.096
(-0.333)
A?e Squared
0.000
(0.088)
-0.009
(-0.951)
0.001
(0.155)
luoer of living Childrei
-0.259
(-2.409)
-0.094
(-0.392)
-0.056
(-0.385)
Priiarv School
-0.452
(-0.111)
-1.443
(-1.316)
-1.023
(-1.418)
Jr. Hign School
-1.502
(-2.311)
-1.533
(-1.266)
-0.691
(-0.908)
Sr. Hign School
-0.186
(-1.212)
-1.211
(-0.991)
-0.198
(-0.981)
Field rforxer Using
1.544
(2.421)
-8.582
(-0.018)
1.650
(2.223)
Santo (uh
0.651
(1.545)
0.114
(0.848)
0.316
(0.705)
Taoin
-0.904
(-2.213)
•1.309
(-1.066)
-0.095
(-1.898)
Custaat
4.510
(1.020)
-15.922
-1.041)
3.111
(0.681)
Ecscation*
CoBtriceotion,M
Leg likeiihood:-326J2; OHL
••
Ounv Variable: Excluded category is no schooling.
Oumv Variable: Excluded category is field rorker not usieg contraception.
Oumv variable: Excluded category is 8anjar Regency.
- Media
CHILD SURVIVAL.pdf
Position: 6091 (1 views)