Machine learning modeling for identifying predictors of unmet need for family planning among married/in-union women in Ethiopia: Evidence from performance monitoring and accountability (PMA) survey 2019 dataset

Unmet need for contraceptives is a public health issue globally that affects maternal and child health. Reducing unmet need reduces the risk of abortion or childbearing by preventing unintended pregnancy. The unmet need for family planning is a frequently used indicator for monitoring family planning programs. This study aimed to identify predictors of unmet need for family planning using advanced machine learning modeling on recent PMA 2019 survey data. The study was conducted using secondary data from PMA Ethiopia 2019 cross-sectional household and female survey which was carried out from September 2019 to December 2019. Eight machine learning classifiers were employed on a total weighted sample of 5819 women and evaluated using performance metrics to predict and identify important predictors of unmet need of family planning with Python 3.10 version software. Data preparation techniques such as removing outliers, handling missing values, handling unbalanced categories, feature engineering, and data splitting were applied to smooth the data for further analysis. Finally, Shapley Additive exPlanations (SHAP) analysis was used to identify the top predictors of unmet need and explain the contribution of the predictors on the model’s output. Random Forest was the best predictive model with a performance of 85% accuracy and 0.93 area under the curve on balanced training data through tenfold cross-validation. The SHAP analysis based on random forest model revealed that husband/partner disapproval to use family planning, number of household members, women education being primary, being from Amhara region, and previously delivered in health facility were the top important predictors of unmet need for family planning in Ethiopia. Findings from this study suggest various sociocultural and economic factors might be considered while implementing health policies intended to decrease unmet needs for family planning in Ethiopia. In particular, the husband’s/partner’s involvement in family planning sessions should be emphasized as it has a significant impact on women’s demand for contraceptives.Author summary: This study aimed to identify key predictors contributing to women’s unmet need for family planning in Ethiopia. Eight machine learning algorithms were trained, and the random forest model was the most accurate with an area under the curve score of 0.93 and an accuracy of 85%. The top ten predictors were identified using the SHAP feature importance method, with husband/partner disapproval, primary education, being from the Amhara region, having previously delivered in health facilities, and middle wealth quantile being key factors. Women living in lower household members also increased the likelihood of unmet need for family planning. Unmet need for family planning can lead to unintended pregnancies, negative health outcomes for the mother and child, and economic and social consequences. It can also affect population growth rate, as women may have more children than they would otherwise choose to have, putting strain on resources and causing economic instability. To effectively address unmet needs, women’s education level, number of households, residence region, and place of delivery must be considered. Emphasizing husband/partner involvement in family planning sessions is crucial, as it significantly impacts women’s demand for contraceptives.