Determinants of infant mortality in Malawi: an analysis to control for death clustering within families

Two logistic binomial models for neonatal mortality (under 1 month) and post-neonatal mortality were used to determine the probability of dying among families in Malawi. Data was obtained from 3043 women aged 15-54 years on 6258 births, which occurred 0-15 years before the survey, from the 1988 Malawi Traditional and Modern Methods of Child Spacing Survey. Mortality included 211 post-neonatal deaths, 147 toddler deaths, and 172 child deaths. Missing information or date of death missing information pertained to 182 reported deaths that were excluded from the analysis. Logistic models were run with the complete sample and the sub-sample and found to have similar results. Findings showed that children born in homes with electricity had 34% lower risk of dying than children born in homes without electricity. Preceding birth interval was unrelated to neonatal mortality. Neonatal mortality rates were significantly higher in Chiradzulu rural area, which was found to have a lower proportion of mothers with five or more years of education. The random term, which was high, suggested a high familial correlation with neonatal mortality risk. Findings showed that families with favorable characteristics living in the Chiradzulu area had a probability of 0.005 of a neonatal death. Low risk families in unfavorable circumstances had lower probabilities of child loss than high risk families with favorable conditions. Significant determinants of post-neonatal mortality were preceding birth interval, maternal education, father's occupation, and geographic area. Women with 9 or more years of education had lower infant mortality risks. Family effects were significant, even after controlling for socioeconomic conditions. The most favorable conditions for child survival were: no preceding child; a preceding birth interval of 19 months or longer; maternal education of 9 or more years; and paternal employment in non-manual work