Background: Annual global estimates of perinatal mortality show Malawi among sub-Saharan Africa with the highest rates. Targeted interventions are required to reduce this mortality. This study aimed to quantify small-scale geographical variations in perinatal mortality, and estimate risk factors associated with perinatal mortality in Mchinji district.
Methods: As part of the 2005-2010 randomised controlled trial conducted in Mchinji district, prospective data from the control arm of the trial was collected on perinatal mortality. A Structured Additive Regression model was applied to account for influence of both individual and contextual factors, and jointly accounting for nonlinear effects of continuous covariates, spatially structured variation, unstructured heterogeneity and fixed effects. Modelling and inference used a fully Bayesian approach.
Results: Factors associated with reduced perinatal mortality were: previous pregnancy; early and consistent use of antenatal care; syphilis test; abdominal examination; pregnancy danger signs advice; skilled birth attendant; normal labour duration; gestation period of at least 9 months; and normal delivery. Perinatals whose mothers had blood test were associated with high probability of dying. Perinatals from mothers between 16 and 40 years had reduced prevalence of dying while those aged less than 16 years and greater than 40 years were associated with higher prevalence of dying. After accounting for all significant covariates, high perinatal mortality was observed in eastern part of the district whereas low perinatal mortality was observed in the western part.
Conclusion: Targeting health interventions to higher risk areas and ensuring universal coverage are promising approaches for promoting equity and reducing perinatal mortality.
Keywords: Full bayesian; MCMC; Perinatal mortality and sub-Saharan Africa; STAR models.
Copyright © 2016 Elsevier Ltd. All rights reserved.