关键词: 妊娠期糖尿病, 盆底功能障碍, 影响因素, 决策树, Logistic回归, 预测模型

Abstract:

Objective　To investigate the value of decision tree and logistic regression model in the prediction of postpartum pelvic floor dysfunction in patients with gestational diabetes mellitus. Methods　Two hundred and four patients with gestational diabetes who gave birth and underwent pelvic floor function screening 42 d after delivery in Northwest Women's and Children's Hospital from September 2022 to September 2024 were selected as the study objects. The clinical data, such as age, pre-pregnancy body mass index (BMI), etc. were collected. According to whether they had postpartum pelvic floor dysfunction, the patients were divided into a pelvic floor dysfunction group and a non-pelvic floor dysfunction group. The clinical data were compared between the two groups. The decision tree model and multivariate logistic regression analysis model were constructed to analyze the risk factors of postpartum pelvic floor dysfunction in the patients with gestational diabetes mellitus. The data were compared between the two groups by independent sample t test and χ² test. A receiver operating characteristic curve (ROC) was drawn to analyze the risk prediction effects of the decision tree model and multivariate logistic regression model. Results　There were statistical differences in age, parity, neonatal weight, second stage of labor, history of pelvic operation, and pelvic floor function rehabilitation exercise between the two groups(all P<0.05). The multivariate logistic regression analysis showed that ≥35 years old (OR=3.702), parity ≥2 times (OR=2.054), neonatal weight > 4 000 g (OR=2.516), second stage of labor > 1 h (OR=7.744), history of pelvic surgery (OR=9.052), and no postpartum pelvic floor rehabilitation exercise (OR=18.504) were considered as the risk factors of postpartum pelvic floor dysfunction in the women with gestational diabetes mellitus (all P<0.05). The decision tree model showed that postpartum pelvic floor function rehabilitation exercise had the strongest correlation with postpartum pelvic floor dysfunction in the patients (χ²=27.886; P<0.001). The ROC analysis showed that the area under the curve (AUC) of the decision tree model for predicting postpartum pelvic floor dysfunction in the patients was 0.939, which was higher than the AUC of the multivariate logistic regression model (0.807). Conclusion　Both the decision tree and logistic regression model have certain predictive value for postpartum pelvic floor dysfunction in patients with gestational diabetes mellitus, and the decision tree model has a higher sensitivity.

Key words: Gestational diabetes mellitus,  , Pelvic floor dysfunction,  , Influencing factors,  , Decision tree,  , Logistic regression,  , Prediction model