Objective:To explore risk factors for delayed onset of lactogenesis in prenatally overweight women and construct risk prediction models based on machine learning for early identification of high-risk individuals.Methods:Convenience sampling was adopted to select 338 prenatally overweight women who delivered in the Obstetrics Departments of four ClassⅢ Grade A hospitals in Hengyang City from October 2023 to June 2024 for the study. Delivery women were randomly divided into training set and test set in the ratio of 7∶3. The survey was conducted with the General Information Questionnaire, Breastfeeding Self-efficacy Scale Short Form, Edinburgh Postnatal Depression Scale, LATCH Scale and Pittsburgh Sleep Quality Index. One-way analysis and LASSO regression were used to screen predictors using delayed onset of lactogenesis as the outcome variable. Risk prediction models were constructed based on three machine learning algorithms of Logistic regression, support vector, and random forest, respectively. The models were tuned by ten-fold cross-validation to filter out the best models.Results:Delayed onset of lactogenesis occurred in 140 of 338 prenatally overweight women, an incidence of 41.4%. Among the three predictive model performances, the random forest model had the highest area under the receiver operating characteristic curve, accuracy, precision, recall, and F1 value. The importance of each predictor was ranked according to the fandom forest algorithm, and in descending order of importance, they were breastfeeding 1 h after the birth of the newborn, number of previous deliveries, age, feeding mode 3 d postpartum, pregnancy complications, mode of delivery, number of breastfeeding 24 h postpartum, and monthly household income.Conclusions:Risk prediction models for delayed onset of lactogenesis in prenatally overweight women are constructed based on three machine learning algorithms, aiming to help provide a scientific basis for clinical healthcare professionals to take relevant decisions.

Objective:To explore risk factors for delayed onset of lactogenesis in prenatally overweight women and construct risk prediction models based on machine learning for early identification of high-risk individuals.Methods:Convenience sampling was adopted to select 338 prenatally overweight women who delivered in the Obstetrics Departments of four ClassⅢ Grade A hospitals in Hengyang City from October 2023 to June 2024 for the study. Delivery women were randomly divided into training set and test set in the ratio of 7∶3. The survey was conducted with the General Information Questionnaire, Breastfeeding Self-efficacy Scale Short Form, Edinburgh Postnatal Depression Scale, LATCH Scale and Pittsburgh Sleep Quality Index. One-way analysis and LASSO regression were used to screen predictors using delayed onset of lactogenesis as the outcome variable. Risk prediction models were constructed based on three machine learning algorithms of Logistic regression, support vector, and random forest, respectively. The models were tuned by ten-fold cross-validation to filter out the best models.Results:Delayed onset of lactogenesis occurred in 140 of 338 prenatally overweight women, an incidence of 41.4%. Among the three predictive model performances, the random forest model had the highest area under the receiver operating characteristic curve, accuracy, precision, recall, and F1 value. The importance of each predictor was ranked according to the fandom forest algorithm, and in descending order of importance, they were breastfeeding 1 h after the birth of the newborn, number of previous deliveries, age, feeding mode 3 d postpartum, pregnancy complications, mode of delivery, number of breastfeeding 24 h postpartum, and monthly household income.Conclusions:Risk prediction models for delayed onset of lactogenesis in prenatally overweight women are constructed based on three machine learning algorithms, aiming to help provide a scientific basis for clinical healthcare professionals to take relevant decisions.