Pooled Prevalence, Spatial Patterns and Determinants of Time to Pregnancy Loss Among Reproductive-Aged Women in East Africa: Bayesian Spatial Frailty Model

Abstract

Background: Pregnancy loss is the terminated pregnancy before the completed pregnancy time. It persists as a neglected concern, overlooked in policy agendas and programs and demanding immediate and focused attention. Despite ongoing efforts to reduce the global burden of pregnancy loss, significant challenges remain, particularly in East Africa. It is therefore essential to understand the spatial patterns and factors that influence differences in survival times for pregnancy loss in order to establish effective policies that can enhance health systems. Objectives: The study aim was to assess the pooled prevalence, spatial patterns and determinants of time to pregnancy loss among reproductive-aged women in East Africa using a secondary data analysis of recent demographic and health survey from 2015 to 2023. Methods: Data was analyzed using R software version 4.4.1. The pooled prevalence of pregnancy loss was obtained using a Bayesian random effects model. Spatial data analysis was conducted to identify geographic variations in pregnancy loss. A spike-and-slab prior was used for variable selection. A Bayesian spatial survival method via an intrinsic conditional autoregressive approach was used to determine factors related to time to pregnancy loss among 169 regions of 9 East African countries. Model comparison was conducted using deviance information criteria, Watanabe-Akaike information criterion, and log pseudo marginal likelihood, while Cox and Snell residuals were used for model diagnostics. Results: The estimated pooled proportion of pregnancy loss among women in East Africa, based on recent DHS data (2015-2023), was 0.14 [95% CrI, 0.1−0.21]. Spatial analysis revealed a clustered distribution of pregnancy loss with significant spatial variation. The Bayesian spatial frailty model identified maternal age, marital status, mothers’ educational level, media exposure, residence, maternal occupation, parity, and antenatal care visits as significant predictors of survival time. After adjusting for known subject-specific covariate effects, spatial dependence in the hazard of pregnancy loss was identified. Conclusion and Recommendation: Pregnancy loss remains a significant public health challenge in East Africa, with significant geographic variation. The spatial analysis and determinates of pregnancy loss can indicate the health policy makers to give special attention to the high-risk regions. Therefore, international organizations, policymakers, and health ministries in East Africa should prioritize resources for high-risk areas to improve maternal and child health.