Direct and indirect associations of five factors with infant mortality.

The relationship between several factors including maternal age, race, education and prior obstetric experience to infant mortality was investigated using vital statistics data from four regionalized networks in 1974 and 1975 and applying the log-linear model. By making a causal assumption, these relationships can be decomposed into a direct component and an indirect component acting through the association of the variables with birth weight. Detailed examination of the interactions between these factors revealed that all of the factors except education could demonstrate a direct component at the 5% significance level in their relationship to neonatal death, while all but birth order had a direct component relating to postneonatal deaths. The strength of the direct components was expressed in the form of odds ratios, leading to a conclusion confirming some earlier reports from these data, but suggesting that births to mothers over 35 years of age have smaller increased neonatal risk relative to those aged 18-34 years than had been reported previously. In addition, the direct component associated with first births does not show a higher risk than those of birth order 2 and 3. Added risk in this group is traceable to the indirect component. The disadvantage associated with having a prior fetal death is only found among low birth orders, with a possible reversal of the pattern in the high parity group. Analysis of postneonatal risks shows an interaction between maternal age and race, in which births to mothers under age 18 had increased risks only for whites.