A Bayesian belief network modelling of household factors influencing the risk of malaria: A study of parasitaemia in children under five years of age in sub-Saharan Africa

Abstract

Studies that focus on integrated modelling of household factors and the risk for malaria parasitaemia among children in sub-Saharan Africa (SSA) are scarce. By using Malaria Indicator Survey, Demographic Health Survey, AIDS Indicator Survey datasets, expert knowledge and existing literature on malaria, a Bayesian belief network (BBN) model was developed to bridge this gap. Results of sensitivity analysis indicate that drinking water sources, household wealth, nature of toilet facilities, mother's educational attainment, types of main wall, and roofing materials, were significant factors causing the largest entropy reduction in malaria parasitaemia. Cattle rearing and residence type had less influence. Model accuracy was 86.39% with an area under the receiver-operating characteristic curve of 0.82. The model's spherical payoff was 0.80 with the logarithmic and quadratic losses of 0.53 and 0.35 respectively indicating a strong predictive power. The study demonstrated how BBN modelling can be used in determining key interventions for malaria control.