2023, Vol. 8, Issue 5, Part B

Kenya is among the nations where the rates of mortality for children under the age of five are alarmingly high. Under-five mortality rate in Kenya, as reported by the 2014 Kenya Demographic Health Survey, stood at 52 deaths per 1000 live births. It is crucial to tackle the significant issue of identifying the factors contributing to the mortality of children under the age of five, as this knowledge can inform health strategies and interventions. One method for determining these factors is by employing regression models that make different assumptions, including assuming the presence of mediation. Mediation analysis is frequently performed to deepen our comprehension of the underlying mechanisms in well-established cause-and-effect relationships. The primary objective of this analysis is to disentangle the indirect effect, which operates through a specific intermediary (or mediator), from the remaining direct effect. It also quantifies the contributions of these effects to the overall impact of the exposure. In most mediation analyses, even when the focus is on a single mediator, it is important to consider the possibility of multiple mediators. Incorporating additional mediators decreases the chances of overlooking alternative pathways that connect the exposure to the outcome. However, the development of mediation models involving multiple mediators is currently constrained or needs more extensive advancement. The Aalen additive hazard model with multiple mediators is used. Aalen additive hazard models are employed in mediation analyses to consider the influence of the disparity in hazard rates. When conducting causal mediation analyses of survival outcomes involving multiple mediators, additive hazard models are utilized with utility demonstrated in an under five child mortality study (UFCM). The study involved investigation of the effect of maternal education on child mortality, in the presence of two mediators: maternal income and maternal health behaviour. We initially conducted one-mediator analyses, to examine the effect of the maternal education on UFCM mediated through the maternal health behaviour. A two-mediator model investigating path-specific effects is then conducted with the exposure being education level(s), two mediators, one being maternal income (M_1 ) and the other mediator being maternal health behaviour (M_2 ) and the outcome being under five child mortality (Y).

The influence of education on UFCM had different pathways in this study. The effect mediated through maternal income M1 and possibly through maternal health behaviour M2, had stronger effects (−0.36) than ∆S→M_2→Y the effect of education on the outcome mediated only through maternal health behaviour M_2 (-0.036) (i.e not through maternal income M1). Health activities and information relating to maternal health behaviour will have a significant effect if maternal income is increased.