Spatial modelling and mapping of antenatal care service utilization in Ethiopia: an analysis of Ethiopian demographic and health survey, 2011

RESEARCH REPORT SUBMITTED TO SCHOOL OF PUBLIC HEALTH, FACILITY OF HEALTH SCIENCE, UNIVERSITY OF WITWATERSRAND IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE IN EPIDEMIOLOGY UNDER DIVISION OF EPIDEMIOLOGY AND BIOSTATISTICS. April 12, 2016, Johannesburg, South AfricaBACKGROUND: Adequate antenatal care (ANC) is an important indicator to improve maternal and their infant health in a given region. However, ANC coverage varies markedly within geographical locations due to different social and health determinants. Evaluating and documenting such variations provide useful information to improve ANC uptake. OBJECTIVE: To identify socio-economic and demographic factors associated with ANC service utilization, to investigate spatial clustering in ANC service utilization and to develop maps of spatial variation in ANC service utilization in Ethiopia. METHODS: Data was drawn from the 2011 Ethiopian Demographic and Health Survey (EDHS). The two aspects considered under ANC service utilization include the time of first ANC visit and the number of ANC visits. Time of first ANC visit was coded as late (0) when the first visit took place after 16 weeks of gestation otherwise early (1). Number of ANC visit was ordered and coded as no visit (0), inadequate (1) for one to three visits and adequate (2) for greater than or equal to four visits. Alternatively, number of ANC visit was coded as no visit (0) and at least one visit (1) for mapping of spatial variation in number of ANC visit. Spatial scan statistical analysis was carried out on enumeration areas (EAs) aggregated data by using SaTScan to investigate significant clusters of time of first ANC visit and number of ANC visit. High and low rates of ANC service utilization clusters were detected. For time of first ANC visit, the cluster with relative risk (RR) less than one indicates the early visit is lower than late. The number of ANC visits ordered in to three categories, due to this RR in every cluster is represented by three different numbers: the first one stands for (no visits), second for (inadequate visits) and third for (adequate visits). The clusters with RR decrease from low valued category (no ANC visits) to high valued category (adequate number of ANC visits) indicate the rate of number of ANC visits decreases in the clusters. Ordinal logistic regression used to investigate factors associated with the number of ANC visits. Bayesian hierarchical spatial logistic models were used to investigate factors associated with time of first ANC visit and develop maps of spatial variation in the time of ANC visit and the number of ANC visits across Ethiopia. RESULTS: The overall prevalence of time of first ANC visit and adequate number of ANC visits were 26.38% and 19.14% respectively. The result revealed significant spatial variation in ANC service utilization. Women in the West [RR=0.33, P<0.0001] and South-West [RR=0.52, P