SPATIAL DISTRIBUTION AND PREDICTORS OF CONTRACEPTIVE DISCONTINUATION AMONG REPRODUCTIVE-AGE WOMEN IN ETHIOPIA USING 2016 DATA: PREDICTION USING GIS AND MACHINE LEARNING

Abstract

Abstract Introduction: Globally, 38% of contraceptive users discontinue the use of contraceptives. In Ethiopia, about 35% of contraceptive users discontinue within twelve months. Discontinuation reduces contraceptive coverage, family planning program effectiveness and contributes to undesired fertility. Hence understanding the spatial distribution and potential predictors of contraceptive discontinuation is crucial to reducing its undesired outcomes. Thus, this study could enable to determine the spatial heterogeneity and determinant factors for contraceptive discontinuation in Ethiopia. Objective: This study aimed to determine the spatial distribution and predictors of contraceptive discontinuation among reproductive-age women in Ethiopia. Methodology: A population-based cross-sectional study was conducted using secondary data analysis from 2016 Ethiopian Demographic Health Survey. ArcGIS Pro version 2.8 was used to handle mapping, hotspot, and spatial autocorrelation analysis. Bernoulli model was used to analyze the purely spatial cluster detection through SaTScan version 10.0.2 software. Eight machine learning algorithms were employed and evaluated using performance metrics. The best-performing model was applied to predict and identify important predictors of contraceptive discontinuation. Furthermore, the most important factors identified through the best predictive model were used to predict contraceptive discontinuation in originally unsampled areas. Finally, association rule mining was applied to discover the relationship between contraceptive discontinuation and its top predictors. Result: Spatial distribution of contraceptive discontinuation was clustered in Ethiopia with a global Moran’s I index value of 0.3(p-value <0.001). Accordingly, enumeration areas in South Wollo, South Gondar, East and West Gojjam zones of Amhara region, and Illubabor zone of Oromia region were detected as hot spot areas. Random Forest was the best predictive model with 68% accuracy and the top ten predictors of contraceptive discontinuation were identified. Association rule mining identified women's age, women’s education level, family size, husband’s desire for children, husband’s education level, and women’s fertility preference as the most frequently associated factors with contraceptive discontinuation.