Scholarlink

Background <p>Facility delivery attended by skilled health professionals is a critical strategy for reducing maternal and neonatal mortality. Despite substantial efforts to expand maternal health services in Ethiopia, a large proportion of births still occur at home. This study aimed to predict facility delivery and identify its associated factors among reproductive-age women in Ethiopia using machine learning approaches.</p> Methods <p>This study used secondary data from the 2019 Performance Monitoring for Action (PMA) Ethiopia cross-sectional household and female survey. A weighted sample of 5,413 women aged 15–49 years who had a recent birth was included in the analysis. Data extraction and preprocessing were conducted using STATA version 17, while machine learning models were implemented using Python version 3.11.5. Seven supervised machine learning algorithms were developed to predict facility delivery. Model performance was evaluated using accuracy, precision, recall, F1 score, and the area under the receiver operating characteristic curve (AUC). Data preparation included feature engineering, data splitting, handling missing values, addressing class imbalance, and outlier detection. SHapley Additive exPlanations (SHAP) were applied to interpret the contribution of important predictors to model predictions.</p> Results <p>Among the evaluated models, the Random Forest algorithm demonstrated the best predictive performance, achieving 74.15% accuracy, 72.58% recall, 76.27% F1-score, and 80.36% precision on test data. The most influential predictors of facility delivery included higher household wealth status, secondary or higher maternal education, television ownership, discussion of delivery location with a partner, partner encouragement for antenatal care attendance, family planning use, maternal age between 20 and 34 years, age at first sexual intercourse of 19 years or older, awareness of emergency contacts, and completion of primary education.</p> Conclusions <p>Interpretable machine learning approaches can provide useful insights for identifying factors associated with facility delivery. The findings highlight key socioeconomic and behavioral characteristics that may inform targeted strategies to improve utilization of facility-based delivery services and support efforts to enhance maternal health outcomes in Ethiopia.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predicting facility delivery and its determinants among reproductive-age women in Ethiopia using machine learning algorithm: evidence from performance monitoring for action Ethiopia 2019 dataset

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月30日

查看全文
  • Siraj Muhidin Degefa,
  • Beriso Alemu Hailu,
  • Naol Gonfa Serbessa,
  • Asmamaw Ketemaw Tsehay,
  • Tigist Tollessa Sedi,
  • Eden Ketema Woldekidan,
  • Mohammedjud Hassen Ahmed,
  • Jibril Bashir Adem,
  • Agmasie Damtew Wale,
  • Habtamu Alganeh Guadie,
  • Mulusew Andualem Asemahagn

摘要

Background

Facility delivery attended by skilled health professionals is a critical strategy for reducing maternal and neonatal mortality. Despite substantial efforts to expand maternal health services in Ethiopia, a large proportion of births still occur at home. This study aimed to predict facility delivery and identify its associated factors among reproductive-age women in Ethiopia using machine learning approaches.

Methods

This study used secondary data from the 2019 Performance Monitoring for Action (PMA) Ethiopia cross-sectional household and female survey. A weighted sample of 5,413 women aged 15–49 years who had a recent birth was included in the analysis. Data extraction and preprocessing were conducted using STATA version 17, while machine learning models were implemented using Python version 3.11.5. Seven supervised machine learning algorithms were developed to predict facility delivery. Model performance was evaluated using accuracy, precision, recall, F1 score, and the area under the receiver operating characteristic curve (AUC). Data preparation included feature engineering, data splitting, handling missing values, addressing class imbalance, and outlier detection. SHapley Additive exPlanations (SHAP) were applied to interpret the contribution of important predictors to model predictions.

Results

Among the evaluated models, the Random Forest algorithm demonstrated the best predictive performance, achieving 74.15% accuracy, 72.58% recall, 76.27% F1-score, and 80.36% precision on test data. The most influential predictors of facility delivery included higher household wealth status, secondary or higher maternal education, television ownership, discussion of delivery location with a partner, partner encouragement for antenatal care attendance, family planning use, maternal age between 20 and 34 years, age at first sexual intercourse of 19 years or older, awareness of emergency contacts, and completion of primary education.

Conclusions

Interpretable machine learning approaches can provide useful insights for identifying factors associated with facility delivery. The findings highlight key socioeconomic and behavioral characteristics that may inform targeted strategies to improve utilization of facility-based delivery services and support efforts to enhance maternal health outcomes in Ethiopia.