Infant and Child Mortality in Nepal by Norman Y. Luther and Shyam Thapa. East-West Center Working Papers, Population and Health Series, No. 105. April 1999. 100 pp.

Abstract

The 1996 Nepal Family Health Survey (NFHS), part of the World Demographic and Health Surveys, collected information on fertility, family planning, and maternal and child health from 8,429 ever-married women aged 15–49. The data were collected during the months of January through June of 1996. Birth histories of women included in the survey provide information for analysis of child characteristics. Basic results from the survey, including analyses of infant and child mortality, have been published in the first report (Ministry of Health 1997a). This report includes measures of levels and trends of mortality before age 5 by 5-year periods preceding the survey, as well as differentials of mortality by selected socioeconomic background characteristics, demographic characteristics, and health care behavior. A second report has focused on selected topics, including early infant survival (Ministry of Health 1997b). However, a major limitation of both of these reports is that they do not identify the relative effects of various explanatory factors "net" of the effects of others.

Here we remove this limitation by controlling for possibly confounding variables in our investigation of the effects of socioeconomic, demographic, and health care characteristics of children and mothers on infant and child mortality. Because many factors associated with variations in infant and child mortality are interrelated, it is important to estimate the "net" effect of each factor. Hazard model analysis allows us to estimate net effects of each characteristic controlling for the effects of other characteristics. The hazard model can be viewed as a regression model of cohort life tables (Cox, 1972).

Because major causes of death differ substantially at different ages, the effects on mortality of factors we examine may be quite different at different ages of children. The hazard models, therefore, are estimated for three age intervals separately: the neonatal period (first month), postneonatal period (exact ages one to 12 months), and childhood period (exact ages 12 to 60 months).

Results from hazard model analysis are transformed into familiar measures of mortality such as neonatal mortality, postneonatal mortality, infant mortality, and child mortality. The effect of a factor will be presented in terms of differentials in mortality by the factor. The effect of education, for example, is shown by presenting estimated neonatal, postneonatal, infant, and child mortality for children with mothers of different education levels, controlling for the effects of other factors such as sex of child, birth order, year of birth, mother's age at childbirth, mother's age at first marriage, mother's ideal numbers of boys and girls, ecological-developmental region, urban-rural residence, ethnicity, husband's educational level, mother's and husband's work status, exposure to media, standard of living index, selected housing characteristics, length of previous birth interval, number of child's deceased older siblings, sex composition of child's surviving older siblings, and selected maternal and child health care characteristics.

The reliability of mortality estimates calculated from retrospective birth histories depends upon the completeness with which child deaths are reported and the extent to which date of birth and age at death are accurately reported and recorded. In these respects generally the 1996 NFHS data is of good quality and would be expected to generate reasonably reliable mortality estimates. The reader is directed to the national report (Ministry of Health 1997a; see especially Appendix C) and to Retherford and Thapa (1997) for additional details regarding the quality of the data.

Mortality in Nepal declined very substantially over the entire period covered by the 1996 Nepal Family Health Survey (NFHS). Relative decline is largest for child mortality followed by postneonatal mortality and neonatal mortality. The gross effects of time period on mortality estimated by hazard models are consistent with the findings based on cohort life tables presented in Section 4.

Our investigation of the relationship between socioeconomic and demographic background characteristics and mortality under age 5 illustrates the importance of introducing controls. This is especially true in sorting out the real importance of birth order and mother's age at childbirth, each of which shows very different unadjusted and adjusted mortality rates. Contrary to what the unadjusted rates would lead us to believe, in Nepal it is neither birth order one nor advanced maternal age per se that carries exceptionally high risk; rather it is young maternal age and high birth order, which are related to birth order one and advanced maternal age, respectively, that carry the real danger, as revealed by the adjusted rates. Our results show especially high estimated net mortality rates of all four types (neonatal, postneonatal, infant, and child mortality) for children born to very young mothers and children of high birth order. Both of these problems could be greatly reduced by successful programmes promoting contraceptive use. Thus it would seem easier to reduce infant and child mortality by targeting young motherhood and large families with a successful contraceptive use programme than affect such background characteristics as mother's education, ecological-developmental region (by a transmigration programme from the mountains to other regions, for example), or certainly ethnicity, which are the next strongest predictors of infant and child mortality. Successful contraceptive use programmes might also reduce the mother's ideal number of girls (daughters) which has a very strong and statistically significant association with all types of mortality. However, very likely the mother's ideal number of girls is simply a very good indicator of mortality rather than having any causation implications.

The length of a child's previous birth interval is an extremely strong predictor of infant and child mortality, perhaps the strongest of any variable in our entire study. At all ages up to five years, a very substantial reduction in the risk of dying is associated with avoidance of previous birth intervals of less than 24 months, as has been found frequently in other studies. But what seems surprising is the further substantial reduction in mortality at all ages up to five years associated with extending the previous birth interval from beyond 24 months to beyond 36 months, and even more so, to beyond 48 months.

The association between whether the child has any dead older siblings (at the time of his/her birth) and mortality up to age five is not so straightforward. There is a very substantial net effect favoring the child with no dead older siblings during the first month of life, but this relationship weakens at ages 1–11 months and is reversed thereafter. The advantage (lower mortality) after one year of age thus associated with having dead older siblings is not easy to explain. Perhaps extra care is being given to guard against a another child death. But then one would expect somewhat the same results for postneonatal mortality as for child mortality because postneonatal mortality also depends on the quality of care given the child, although not as much as child mortality.

Our study of socioeconomic and demographic background characteristics in section 5 shows that female children are at significantly greater risk of dying than male children after one year of age, although not before. Our further investigation of the effect of the interplay (interaction) between the child's sex and the sex composition of the child's surviving older siblings in section 6 would suggest that inferior care is being given to female children at 1–4 years of age, the ages at which such care has the greatest bearing on survival. This is likely to be a manifestation of son preference which is well known to exist in South-central Asia.

The very strong effects of the length of previous birth interval on all types of mortality hold up under the addition of the survival of previous child variable and mother's ever use of contraceptives variable as controls. The same is true for the effects of child's sex and sex composition of surviving older siblings. This once again suggests that female children are receiving inferior care after the age of one year.

The survival of previous child variable and mother's ever use of contraceptives variable were added as control variables because they are potentially the best proxies available for length of breastfeeding of the previous child and mother's use of contraceptives during the index child's previous birth interval, respectively, given the restrictions of the data. However, as already mentioned, they did not alter the previous results when added as controls. On the other hand, the two variables are of interest in their own right. In particular, mother's ever use of contraceptives is a strong and independent predictor of infant and child mortality when the other variables are controlled.

Several of the maternal and child health care variables stand out as strong predictors, despite the small size of the sub-sample for which they are defined that inhibits statistical significance. For predicting both neonatal and postneonatal mortality, number of antenatal care visits is the strongest predictor and whether mother has received two or more tetanus immunizations during pregnancy ranks next. However, as a predictor of neonatal mortality only, height of mother is easily the strongest, and whether mother has ever used family planning is stronger than whether she has received at least two tetanus immunizations during pregnancy; in addition, whether the mother had excessive bleeding during delivery is slightly stronger. As a predictor of postneonatal mortality only, postpartum care is slightly stronger than whether the mother has received at least two tetanus immunizations during pregnancy.

The relationship between maternal and child care characteristics and mortality under age 5 has one aspect particularly unique to this study. Although gross effects of several of the variables we examined are not particularly substantial and are not statistically significant, partly as a result of the small sub-sample size, in many cases the effects hold up quite well under the introduction of controls—in fact, even increase in some cases. A case in point is height of mother, the strongest predictor of neonatal mortality.

Our results for the maternal and child health care model show high net mortality rates of both types (neonatal and postneonatal mortality) for children born to mothers making no antenatal care visits and not receiving the recommended tetanus dosage during pregnancy; high net neonatal mortality rates for children of mothers who are of small stature (height), who had excessive bleeding during delivery (but data reliability for this variable is questionable), who were assisted only by a friend/relative/other at delivery, and who had never used family planning; and high net postneonatal mortality rates for children of mothers who delivered at home, who had no assistance at delivery, who wanted their pregnancy later, who had long labor during delivery (but data reliability for this variable is questionable), who are of small stature, and who received no postpartum care except possibly by non-medical personnel.

Most of these problems could be greatly reduced by successful maternal health programmes. Indeed, our results indicate that targeting programmes designed to improve maternal nutrition and health will have the most effect on reducing infant mortality. In particular, there should be special emphasis on promoting and providing antenatal care, tetanus immunizations during pregnancy, postpartum care, and assistance at delivery by a traditional birth attendant (TBA). In view of our findings, effective implementation of any of these interventions should substantially reduce infant and child mortality, given the fact that relatively few mothers

In this regard, it is crucial to mention that very generally lack of access to maternal and child health care services may be a major factor in their limited use. A serious limitation to this study has been the inability to control for access to these services, due to the unavailability of information on the subject from the 1996 NFHS. Thus it is highly recommended that future surveys collect data on access to maternal and child health care services.

In any case, the findings in this study are useful in identifying children who are likely to experience high levels of infant and child mortality. Family health programmes should concentrate their efforts on families with characteristics associated with high infant and child mortality, as identified in this and other studies.