Infant and Child Mortality
Infant and Child Mortality
Infant and Child Mortality
Macleod W. Mwale
This chapter reports on levels, trends, and differentials in infant and child mortality based on
the 2004 MDHS. The information on infant and child mortality is relevant to evaluating the pro-
gress of health programmes and in monitoring the current demographic situation. In addition, the
data can be used to identify subgroups of the population that have high mortality risks.
The data for the calculation of mortality rates are collected in the reproduction section of the
Women’s Questionnaire. The section begins with aggregate questions about the total number of
sons and daughters who live with the mother, the number who live elsewhere, and the number who
have died. Then a detailed birth history is administered. For each live birth, information is obtained
on the child’s name, date of birth, sex, whether the birth was single or multiple, and survivorship
status. For living children, information about his or her age at last birthday and whether the child
resides with his or her mother is obtained. For children who had died, the respondent is asked to
provide the age at death.
8.1 DEFINITIONS
The mortality rates presented in this report are defined as follows:
Neonatal mortality (NN): the probability of dying within the first month of life
Postneonatal mortality PNN): the difference between infant and neonatal mortality
Infant mortality (1q0): the probability of dying before the first birthday
Child mortality (4q1): the probability of dying between the first and the fifth
birthday
Under-five mortality (5q0): the probability of dying between birth and the
fifth birthday
All rates are expressed per 1,000 live births, except for child mortality, which is expressed per
1,000 children surviving to 12 months of age.
Population censuses and demographic surveys are the major sources of mortality data in Ma-
lawi, as in most developing countries. Vital registration is another potential source of mortality data.
In Malawi, however, the vital registration data are incomplete in coverage and unrepresentative of
the population. Mortality data from the Health Management Information System (HMIS) is not a
suitable basis for the calculation of mortality rates from a population perspective because the system
is facility-based and does not include data on deaths that occur outside the facilities. Given these cir-
cumstances, birth history data from surveys provide the most reliable estimates of infant and child
mortality for Malawi.
Since only surviving women age 15-49 are interviewed, no data are available for the children
of women who have died. In this case, mortality estimates will be biased if the mortality experience
of children born to surviving and nonsurviving women differs. Of course, any method of estimating
childhood mortality rates that relies on retrospective reporting of events by mothers is susceptible to
bias from this source. The higher the level of adult female mortality and the longer ago the time pe-
riods for which mortality is estimated, the greater is the potential for bias.
Another methodological constraint arises from the fact that women older than age 49 at the
time of the survey are not interviewed and thus cannot contribute information on the exposure and
deaths of their children for periods preceding the survey. This censoring of information and the re-
sulting potential for bias becomes more severe as mortality estimates are made for time periods more
distant prior to the survey. To reduce the effect of these methodological limitations, estimation of
infant and child mortality in this report is restricted to the period 15 years prior to the survey.
The 2004 MDHS Women’s Questionnaire includes two sections on maternal and child
health, in which data are collected on antenatal, delivery, and postnatal care of the mother for recent
births and on many health and nutrition issues for these children (see Chapters 9 and 10). These sec-
tions of the questionnaire must be administered for each birth which occurs after some cut-off date,
typically set to January of the fifth calendar year prior to a survey. In the case of the 2004 MDHS,
the cut-off date was January 1999.
Interviewers in DHS surveys can lessen their workload by recording births that actually occur
after the cut-off date as occurring before that date. This type of birth transference occurs in many
DHS surveys. In the case of the 2004 MDHS, the occurrence of birth transference can be detected
by inspecting the reported number of births in each calendar year before and after the cut-off date
for the health sections. Appendix Table C.4 shows the relevant data. Substantial misreporting of
dates of birth is evident in terms of the calendar year pattern of reported events: 1,575 total births
for 1999 and 2,143 births for 1998 (an increase of 36 percent). Misreporting of dates of birth for
nonsurviving children is even more severe: 233 for 1999 and 424 for 1998 (an increase of 82 per-
cent).
In terms of mortality analysis, what is important is the extent to which this birth transference
distorts the time period in which child deaths occur. To the extent that birth transference results in a
Misreporting age at death can distort the age pattern of mortality. Of particular concern is
the rounding of reported ages at death so that some deaths which actually occur in late infancy are
reported as deaths at one year of age. This type of misreporting would tend to underestimate infant
mortality rates and overestimate child mortality rates. To avoid this problem, interviewers in DHS
surveys are instructed to collect age-at-death data in terms of months of age for children that die af-
ter the first month of life but before two years of age. If a respondent reports the age at death as age
one, the interviewer must probe to determine the number of months that the child lived, being par-
ticularly careful to determine if the child died before or after the first birthday. This procedure of
data collection is designed to minimise the misreporting of age at death and, if digit preference oc-
curs in reported ages at death, it will be obvious from a frequency distribution of deaths by age in
months.
Appendix Table C.6 shows reported deaths by age at death in months (0 through 23 months
of age) and the number of deaths reported as occurring at age one year.2 For the 15-year period im-
mediately preceding the 2004 MDHS, the number of deaths reported at one year of age (422) ex-
ceeds the total number reported at 12 through 23 months of age (403), indicating that interviewers
did not follow standard DHS procedures and making it impossible to assess age at death misreport-
ing by inspection of the distribution of deaths by months of age.
However, the possibility of misreporting late infant deaths as deaths at one year of age can be
indirectly assessed by comparison of the pattern of mortality between the first and the fifth birthday
from the three DHS surveys conducted in Malawi (1992, 2000, and 2004). In each of the three sur-
veys, the age pattern of mortality is similar, with infant mortality rates exceeding child mortality rates
by between 10 and 24 percent. The absence of a significant change in the age pattern of mortality
over the three surveys suggests that, relative to the earlier surveys, substantial age at death misreport-
ing did not occur in the 2004 MDHS.
Underreporting of the births of deceased children (and their subsequent deaths) is always a
concern when collecting birth histories of women. The women may not wish to report such sad
events, and interviewers may fail to record some of these events for the five-year period preceding the
survey in order to avoid asking questions contained in the maternal and child health sections of the
questionnaire.
1
The extent to which the time trend of mortality is distorted by birth transference could be investigated by more detailed
analysis.
2
The number of deaths at one year of age should be minimal in DHS surveys because of the DHS procedure of probing
to determine age at deaths in months when a respondent initially reports one year as the age at death.
The assessment of data quality has found that standard DHS procedures were not followed
in the collection of age-at-death data; that birth dates were misreported (especially in the case of
non-surviving children), resulting in the transference of births out of the five-year period immedi-
ately preceding the survey; and that the ratio of neonatal to infant mortality is unexpectedly lower
for the five-years preceding the survey than for earlier time periods. For these reasons the mortality
estimates from the 2004 MDHS must be interpreted with caution.
During the 15-year period preceding the survey, the estimates indicate that under-five mor-
tality has declined by 30 percent (from 190 deaths per 1,000 to 133 per 1,000). Infant mortality de-
clined by 27 percent (from 104 per 1,000 to 76 per 1,000). Neonatal mortality, however, declined
by 36 percent (from 42 per 1,000 to 27 per 1,000).
Neonatal, postneonatal, infant, child, and under-five mortality rates for five-year periods preceding the
survey, Malawi 2004
The fact that the largest age-specific decline in mortality occurs in the neonatal period is in-
consistent with the pattern of decline usually observed in developing countries. The usual pattern is
greater decline in postneonatal mortality and child mortality than in neonatal mortality, because
some of the causes of neonatal mortality (preterm delivery, injury at delivery, and congenital mal-
formations) are the last to be alleviated in a developing country. Thus it is possible that births ending
in neonatal deaths were underreported for the period immediately preceding the survey, as is sug-
gested in the data quality assessment in Section 8.3.3.
Urban mortality rates are generally lower than rural rates; the under-five mortality rate is 116
per 1,000 in urban areas compared to 164 per 1,000 in rural areas. Comparing the three regions, the
Northern Region has lower under-five mortality (120 per 1,000 live births), than either the Central
(162 per 1,000) or the Southern Regions (164 per 1,000). Similarly, the infant mortality rate is low-
est in the Northern Region (82 per 1,000), compared with either the Central Region (90 per 1,000)
or the Southern Regions (98 per 1,000). These regional differences in mortality were also observed
in the 1992 MDHS and the 2000 MDHS.
Table 8.2 also presents childhood mortality rates for 10 oversampled districts. Under-five
mortality is lowest in Mzimba (112 per 1,000) and Machinga (130 per 1,000) and is highest in Mu-
lanje (221 per 1,000), Kasungu (192 per 1,000), and Thyolo (187 per 1,000). For infant mortality,
the lowest rates are found in Lilongwe (73 per 1,000) and Machinga (78 per 1,000), while the high-
est rates are also observed in Mulanje (145 per 1,000), Thyolo (119 per 1,000), and Kasungu (117
per 1,000).
The 2004 MDHS shows the same relationship between mother’s education and child sur-
vival as the 2000 MDHS. For every age interval, higher levels of education are generally strongly as-
sociated with lower mortality risks. The same is true for the wealth index.
Neonatal, postneonatal, infant, child, and under-five mortality rates for the 10-year period
preceding the survey, by background characteristic, Malawi 2004
As is the case in most populations, male children are more likely to die before reaching the
age of five (166 per 1,000 live births) than female children (149 per 1,000).
The mother’s age at birth is also associated with a child’s chances of survival. Children born
to younger mothers (under 20 years of age) and older mothers (40 years and older) have higher mor-
There is a strong association between the length of the preceding birth interval and mortal-
ity. Under-five mortality of children born following a short birth interval (less than two years) is 67
percent greater than for children born after an interval of 2 years and 162 percent greater than for
children born after an interval of 4 years. This relative mortality disadvantage of children born after a
short birth interval is even more pronounced during the neonatal period.
In the 2004 MDHS, mothers were also asked their perception of the size of their child at
birth for births occurring in the five years preceding the survey. The findings indicate children per-
ceived by their mothers to be small or very small were much more likely to die in the first year of life
(121 per 1,000 live births) than those perceived as average or large in size (65 per 1,000 live births).
Neonatal, postneonatal, infant, child, and under-five mortality rates for the 10-year
period preceding the survey, by demographic characteristics, Malawi 2004
na = Not applicable
1
Computed as the difference between the infant and neonatal mortality rates
2
Excludes first-order births
3
Rates for the five-year period before the survey
There is no consistent relationship between levels of mortality and the first two empower-
ment indicators: participation in household decisionmaking and number of reasons justifying a
woman’s refusal to have sex with her husband. However, there does appear to be a relationship in
the case of attitude towards wife beating. For example, among women reporting fewer reasons justi-
fying wife beating (i.e., more empowered women) under-five mortality is lower (approximately 150
per 1,000) than among women reporting more reasons justifying wife beating (approximately 180
per 1,000).
Neonatal, postneonatal, infant, child, and under-five mortality rates for the 10-year period preced-
ing the survey, by women's status indicators, Malawi 2004