妊娠期尿失禁症状早期预警模型的构建研究

A study on the development of early warning model of stress urinary incontinence during pregnancy

目的:了解妊娠期尿失禁的发生现状,建立妊娠期尿失禁早期预警模型,并检验其效能.方法:对707名初产妇进行回顾性调查.完成社会人口学调查表、危险因素评估量表及尿失禁评定量表.并采用决策树算法进行统计学分析.结果:①当孕产妇符合无尿失禁家族史、有妊娠纹出现、便秘的条件时,发生尿失禁的患者占该节点构成的63.2%;当孕产妇符合无尿失禁家族史、有妊娠纹出现、不便秘、孕前BMI水平≤23.960、孕期每周1-2次盆底肌肉锻炼、每周吃水果1-3天或4-6天、有流产史的条件时,发生尿失禁的患者占该节点构成的58.8%;②危险因素的重要性排序为:便秘、孕前BMI水平、尿失禁家族史、周进食水果量、出现妊娠纹、流产史、孕期盆底肌肉锻炼的强度、孕前体育锻炼的强度、孕前盆底肌肉锻炼的强度、呕吐、受教育水平、日饮水量.③此模型准确度为76.7%,灵敏度为84.8%,ROC曲线下面积达0.767 (95%CI:0.719 - 0.814),P<0.001.交叉验证的风险估计为29.8%,P<0.001.结论:基于决策树算法的妊娠期尿失禁早期预警模型具有较准确的预测能力.在本地区的人群中,其综合预测准确性较优.

Objective: To understand the occurrence of urinary incontinence during pregnancy, to establish an early warning model for urinary incontinence during pregnancy, and to test its effectiveness. Methods: A retrospective survey of 707 primiparas was conducted. The primiparas were asked to complete a questionnaire included demographic information, risk factor evaluated scale and the International Consultation on Incontinence Questionnaire Short Form(ICI-Q-SF). Results: ① If the primipara have this condition of without urinary incontinence family history, with stretch marks and constipation, the incidence rate of urinary incontinence patients accounted for 63.2% of the node. And when the primiparas have this condition of without urinary incontinence family history, with stretch marks, no constipation, pre-pregnancy BMI ≤ 23.96, pelvic floor muscle exercise during pregnancy(1-2 times per week), who eat fruit 1-3 days or 4-6 days a week, with abortion history, the incidence rate of urinary incontinence patients accounted for 58.8% of the node. ② The importance of risk factors were as follows: constipation, pre-pregnancy BMI level, family history of urinary incontinence, weekly fruit consumption, stretch marks, history of abortion, strength of pelvic floor muscle exercise during pregnancy, physical exercise before pregnancy, pelvic floor muscle exercise before pregnancy, vomiting, education level, daily water intake. ③ This model has an accuracy of 76.7%, a sensitivity of 84.8%. The area under the ROC curve was 0.767(95% CI: 0.719-0.814), P<0.001. The risk of cross-validation was estimated at 29.8%, P<0.001. Conclusion: The early warning model of urinary incontinence in pregnancy based on the decision tree algorithm has more accurate predictive ability. Among the population in the region, the overall forecast accuracy is better.