Objective:To construct an early warning model of postpartum urinary retention risk after painless delivery of primipara by using machine learning algorithm,and to find the best effective early warning model,so as to provide scientific basis for early and accurate identification of high-risk groups in clinical practice.Methods:This study was a single-center retrospective study,80 primipara who delivered painlessly in Tingzhou Hospital of Fujian Province from July 2021 to June 2024 were included,they were divided into urinary retention group(18 cases)and non urinary retention group(62 cases)according to whether there was urinary retention after delivery.General data between two groups were compared,Univariate and Multivariate logistic regression were used to screen for influencing factors,three machine learning algorithms:Random Forest,Support Vector Machine,and Logistic Regression were used to construct an early warning model,the area under the receiver operating characteristic curve(ROC-AUC),accuracy,sensitivity,and specificity were used as performance evaluation indicators to evaluate the predictive performance of the model.Results:Univariate analysis showed that age,body mass index(BMI),gestational week,length of the second stage of labor,anesthetic dose,lateral episiotomy were associated with postpartum urinary retention(P＜0.05);Multivariate logistic regression identified BMI ≥ 28 kg/m2(OR=3.210,95％CI:1.450-7.090),length of the second stage of labor ≥ 2 hours(OR=2.890,95％CI:1.230-6.810),anesthetic dose≥ 15 mL(OR=3.560,95％CI:1.670-7.620),and lateral episiotomy(OR=2.540,95％CI:1.120-5.780)as independent risk factors.After comprehensive evaluation of various indicators,the random forest model has the best predictive performance.Conclusion:The risk warning model constructed based on machine learning has good predictive performance,and the random forest model performs the best,which can provide effective support for early clinical intervention.

Objective:To construct an early warning model of postpartum urinary retention risk after painless delivery of primipara by using machine learning algorithm,and to find the best effective early warning model,so as to provide scientific basis for early and accurate identification of high-risk groups in clinical practice.Methods:This study was a single-center retrospective study,80 primipara who delivered painlessly in Tingzhou Hospital of Fujian Province from July 2021 to June 2024 were included,they were divided into urinary retention group(18 cases)and non urinary retention group(62 cases)according to whether there was urinary retention after delivery.General data between two groups were compared,Univariate and Multivariate logistic regression were used to screen for influencing factors,three machine learning algorithms:Random Forest,Support Vector Machine,and Logistic Regression were used to construct an early warning model,the area under the receiver operating characteristic curve(ROC-AUC),accuracy,sensitivity,and specificity were used as performance evaluation indicators to evaluate the predictive performance of the model.Results:Univariate analysis showed that age,body mass index(BMI),gestational week,length of the second stage of labor,anesthetic dose,lateral episiotomy were associated with postpartum urinary retention(P＜0.05);Multivariate logistic regression identified BMI ≥ 28 kg/m2(OR=3.210,95％CI:1.450-7.090),length of the second stage of labor ≥ 2 hours(OR=2.890,95％CI:1.230-6.810),anesthetic dose≥ 15 mL(OR=3.560,95％CI:1.670-7.620),and lateral episiotomy(OR=2.540,95％CI:1.120-5.780)as independent risk factors.After comprehensive evaluation of various indicators,the random forest model has the best predictive performance.Conclusion:The risk warning model constructed based on machine learning has good predictive performance,and the random forest model performs the best,which can provide effective support for early clinical intervention.