Scholarlink

Background <p>Exclusive breastfeeding (EBF) during the first 6&#xa0;months is globally recommended for optimal maternal and child health. Nevertheless, adherence to this recommendation remains suboptimal in China. This study aimed to identify predictors of 6-month EBF among mothers in Jiangsu Province by developing an explainable machine learning (ML) model within a prospective cohort design.</p> Methods <p>Between August 2022 and March 2023, postpartum women were recruited through multistage random sampling across hospitals of different levels. Data were collected via structured discharge interviews and three follow-up calls using validated instruments. Least Absolute Shrinkage and Selection Operator (LASSO) regression was applied for feature selection. Four ML algorithms, including Extreme Gradient Boosting (XGBoost), Random Forest, Decision Tree, and Logistic Regression, were compared using tenfold cross-validated area under the receiver operating characteristic curve (AUC) in the training set. The best-performing algorithm was then retrained on the full training set and evaluated in an independent validation set. SHapley Additive exPlanations (SHAP) was applied to enhance interpretability.</p> Results <p>A total of 374 mothers completed follow-ups. Fewer than half sustained EBF for 2&#xa0;months, about one-third for 4&#xa0;months, and only 12.3% for 6&#xa0;months. XGBoost showed the highest cross-validated performance (mean AUC = 0.75). After retraining, the XGBoost model achieved an AUC of 0.999 in the full training set and 0.853 in the validation set. SHAP analysis identified the most influential predictors in the following order: breastfeeding (BF) intention, subjective norm, perceived control, BF attitude, BF knowledge, maternal education, and exposure to BF education.</p> Conclusions <p>Sustaining EBF for 6&#xa0;months remains challenging. The XGBoost model, interpreted using SHAP, demonstrated acceptable internal performance. It also yielded exploratory yet informative insights into factors influencing 6-month EBF. These findings generate preliminary evidence that may inform locally relevant EBF support efforts and contribute to the growing body of data-driven EBF studies. External validation is required before considering broader applicability.</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

Explainable machine learning model to predict 6-month exclusive breastfeeding: a prospective cohort study in Jiangsu, China

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月26日

查看全文
  • Qian Wu,
  • Chintana Wacharasin,
  • Yan Tang

摘要

Background

Exclusive breastfeeding (EBF) during the first 6 months is globally recommended for optimal maternal and child health. Nevertheless, adherence to this recommendation remains suboptimal in China. This study aimed to identify predictors of 6-month EBF among mothers in Jiangsu Province by developing an explainable machine learning (ML) model within a prospective cohort design.

Methods

Between August 2022 and March 2023, postpartum women were recruited through multistage random sampling across hospitals of different levels. Data were collected via structured discharge interviews and three follow-up calls using validated instruments. Least Absolute Shrinkage and Selection Operator (LASSO) regression was applied for feature selection. Four ML algorithms, including Extreme Gradient Boosting (XGBoost), Random Forest, Decision Tree, and Logistic Regression, were compared using tenfold cross-validated area under the receiver operating characteristic curve (AUC) in the training set. The best-performing algorithm was then retrained on the full training set and evaluated in an independent validation set. SHapley Additive exPlanations (SHAP) was applied to enhance interpretability.

Results

A total of 374 mothers completed follow-ups. Fewer than half sustained EBF for 2 months, about one-third for 4 months, and only 12.3% for 6 months. XGBoost showed the highest cross-validated performance (mean AUC = 0.75). After retraining, the XGBoost model achieved an AUC of 0.999 in the full training set and 0.853 in the validation set. SHAP analysis identified the most influential predictors in the following order: breastfeeding (BF) intention, subjective norm, perceived control, BF attitude, BF knowledge, maternal education, and exposure to BF education.

Conclusions

Sustaining EBF for 6 months remains challenging. The XGBoost model, interpreted using SHAP, demonstrated acceptable internal performance. It also yielded exploratory yet informative insights into factors influencing 6-month EBF. These findings generate preliminary evidence that may inform locally relevant EBF support efforts and contribute to the growing body of data-driven EBF studies. External validation is required before considering broader applicability.