Frontiers in public health · 2026
Identifying early key influencing factors of positive results in the early screening for postpartum depression with interpretable machine learning.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Lin Shusen, Wang Yan, Han Xuefei, Wang Xiaojing, Zhang Yanxia, Xu Wanyu, Chen Meiyu, Xu Xiaobing, Li Huawei
Аннотация
OBJECTIVE: Based on social ecosystem theory, this study explores the early key factors associated with postpartum depressive symptoms from multiple dimensions, and analyzes them using interpretable machine learning algorithms.
METHODS: This was a short-term longitudinal study. Using the convenience sampling method, pregnant women who were hospitalized in the obstetrics departments of two tertiary grade-A hospitals in Shandong Province, China, from July 2023 to January 2025 and met the inclusion and exclusion criteria were selected as the research subjects. Based on social ecosystem theory, candidate factors were discussed in expert meetings, and 53 candidate factors were finalized for the survey. After feature selection using recursive feature elimination combined with 10-fold cross-validation, early risk factors for EPDS screening-positive postpartum depressive symptoms were identified using logistic regression. The model construction was carried out using Python 3.12 to build 9 machine learning algorithm models and conduct internal and external validations. After selecting the best model based on various evaluation indicators, SHAP (Shapley Additive Explanations) analysis was conducted within the best model to obtain the global interpretability analysis.
RESULTS: After recursive feature elimination and logistic regression, six early risk factors for EPDS screening-positive postpartum depressive symptoms were identified: whether there was work pressure during maternity leave, whether forced eating occurred due to breastfeeding, the relationship between the mother and her husband, whether weight gain during pregnancy caused distress, the current sleep situation of the mother, and the degree of social support. After comparing the internal and external validation indicators of the 9 machine learning algorithm models, the LightGBM model was found to be the best model. Finally, the SHAP analysis was presented in the form of swarm plots and Polar Plots, which demonstrated the global feature importance ranking within the best model, the influence direction and distribution of each early risk factors, as well as the interaction pattern between feature values and their impacts.
CONCLUSION: This study combines the social ecosystem theory and interpretable machine learning algorithms to explore the early risk factors for EPDS screening-positive postpartum depressive symptoms, which can provide a reference basis for the subsequent formulation of early risk management plans.
METHODS: This was a short-term longitudinal study. Using the convenience sampling method, pregnant women who were hospitalized in the obstetrics departments of two tertiary grade-A hospitals in Shandong Province, China, from July 2023 to January 2025 and met the inclusion and exclusion criteria were selected as the research subjects. Based on social ecosystem theory, candidate factors were discussed in expert meetings, and 53 candidate factors were finalized for the survey. After feature selection using recursive feature elimination combined with 10-fold cross-validation, early risk factors for EPDS screening-positive postpartum depressive symptoms were identified using logistic regression. The model construction was carried out using Python 3.12 to build 9 machine learning algorithm models and conduct internal and external validations. After selecting the best model based on various evaluation indicators, SHAP (Shapley Additive Explanations) analysis was conducted within the best model to obtain the global interpretability analysis.
RESULTS: After recursive feature elimination and logistic regression, six early risk factors for EPDS screening-positive postpartum depressive symptoms were identified: whether there was work pressure during maternity leave, whether forced eating occurred due to breastfeeding, the relationship between the mother and her husband, whether weight gain during pregnancy caused distress, the current sleep situation of the mother, and the degree of social support. After comparing the internal and external validation indicators of the 9 machine learning algorithm models, the LightGBM model was found to be the best model. Finally, the SHAP analysis was presented in the form of swarm plots and Polar Plots, which demonstrated the global feature importance ranking within the best model, the influence direction and distribution of each early risk factors, as well as the interaction pattern between feature values and their impacts.
CONCLUSION: This study combines the social ecosystem theory and interpretable machine learning algorithms to explore the early risk factors for EPDS screening-positive postpartum depressive symptoms, which can provide a reference basis for the subsequent formulation of early risk management plans.