This paper investigates the association of early childhood mortality (between birth and second birthday) with ethnicity in Nepal, based on data from the 1976 Nepal Fertility Survey, which was part of the World Fertility Survey. The approach is through a series of hazard models, which incorporate ethnicity, year of birth, mother's illiteracy, father's illiteracy, rural-urban residence, region, sex, maternal age, survival of previous birth, previous birth interval, and breast-feeding as covariates. Ethnic differentials in early childhood mortality are not explained by the other socioeconomic and demographic covariates, except for a modest effect of illiteracy, but the remaining covariates explain a great deal of variability in early childhood mortality itself. Analysis using an improved specification of breast-feeding as an age-varying covariate indicates, on average, that breast-feeding, relative to not breast-feeding, reduces age-specific mortality risks during the first 2 years of life by 76%, a very large effect.
PIP: Hazard models were fitted to data from the 1976 Nepal Fertility Survey (NFS) to investigate ethnic differentials in early childhood mortality between birth and 2nd birthday. The NFS included 5940 ever-married women aged 15-49. In addition to data on early childhood mortality, the survey collected considerable information on demographic and socioeconomic factors relevant to childhood mortality. Biomedical and environmental data were not collected. The births in the analysis were restricted to last and next-to last births of order 2 or higher occurring during 1962-76. A table shows the results of fitting 5 hazard models. Model I incorporates only ethnicity and year of birth as covariates. The full model II includes ethnicity and year of birth and also socioeconomic and demographic covariates. Relative to model II, model III deletes the ethnicity covariates, model IV deletes the socioeconomic covariates, and model V deletes the demographic covariates. Comparison of the 5 models indicates how the effects of ethnicity of early childhood mortality are affected by each block of covariates and by all covariates simultaneously. Model I shows that ethnicity, with only year of birth controlled, has a considerable mortality effect, as measured by relative risk. Model II, the complete model, shows that the introduction of socioeconomic and demographic covariates makes little difference in the relative risks by ethnicity. Comparison of model II with model I shows that relative risks increase only slightly for 3 ethnic groups and decrease only slightly for 4, and that levels of statistical significance of ethnicity coefficients are reduced only marginally. Thus, the socioeconomic and demographic covariates contribute little to explaining ethnic differentials in early childhood mortality. With Model III, except for the now significant effects of illiteracy of the mother and, to a lesser extent of the father, which become somewhat stronger, relative risks for the remaining covariates are changed hardly at all by this deletion. Relative risks for the various ethnic groups increase slightly with Model IV, indicating that socioeconomic factors explain some of the effects of ethnicity on early childhood mortality. In model V, deleting demographic covariates from model II, relative risks for the ethnic groups tend to decrease slightly. The average effect of breastfeeding is huge. The early childhood mortality risk for breastfed children is only 24% of that for children who are not breastfed.