Supplementary Material for: Development of Predictive Risk Models of Postpartum Stress Urinary Incontinence for Primiparous and Multiparous Women

Objectives: To develop risk predictive models of postpartum stress urinary incontinence (SUI) for both primiparous and multiparous women. Materials and Methods: From July 2016 to July 2017, 815 singleton pregnant women without incontinence before pregnancy who were 18 years or older and admitted to 2 hospitals in Shenzhen, China, were enrolled. Pregnancy-related data were collected at enrollment. Delivery information was obtained from electronic medical records. Telephone follow-up was conducted to investigate SUI at 6 weeks postpartum. Multivariable logistic regression analyses using stepwise selection were used to establish predictive models for postpartum SUI for all women, and separately for primiparous and multiparous. Internal validation of the models was performed with discrimination and calibration using a bootstrapping (1,000 resampling) method. Results: The analysis included 727 participants. The prevalence of postpartum SUI was 15.96% (116/727), 12.5% (49/393) for primiparous women and 20.1% (67/334) for multiparous women, with a significant difference between them (p = 0.008). For primiparous women, the predictive postpartum SUI model included age, abortion/miscarriage history, SUI during pregnancy, and mode of delivery. For multiparous women, pre-pregnancy BMI, abortion/miscarriage history, SUI during pregnancy, and mode of delivery were included in the model. There was satisfactory calibration between the models’ predicted probability of postpartum SUI and the observed probability for both primiparous and multiparous women (Hosmer-Lemeshow test, p = 0.390 for primiparous and 0.364 for multiparous women). The optimism-corrected C-statistic of the models by bootstrapping stepwise was 0.763 (95% confidence interval [CI]: 0.693–0.833) for primiparous women and 0.783 (95% CI: 0.726–0.841) for multiparous women. Conclusion: We developed predictive models of postpartum SUI for both primiparous and multiparous women. This approach may provide a useful tool for high-risk prediction of postpartum SUI before and after delivery.

研究目标：构建针对初产妇与经产妇的产后压力性尿失禁（postpartum stress urinary incontinence, SUI）风险预测模型。材料与方法：2016年7月至2017年7月，纳入中国深圳2家医院收治的815名妊娠前无尿失禁病史、年龄≥18岁的单胎妊娠女性。入组时收集妊娠相关临床资料，从电子病历中提取分娩相关信息，并通过电话随访调查产后6周时的SUI发生情况。采用逐步变量筛选法构建多因素logistic回归模型，分别针对全部研究对象、初产妇及经产妇建立产后SUI预测模型。采用bootstrap重抽样法（bootstrap，1000次重采样）对模型进行内部验证，通过区分度与校准度评估模型性能。结果：最终纳入727名研究参与者。产后SUI总体患病率为15.96%（116/727），其中初产妇患病率为12.5%（49/393），经产妇为20.1%（67/334），组间差异具有统计学意义（p=0.008）。针对初产妇的产后SUI预测模型纳入年龄、流产/自然流产史、妊娠期SUI及分娩方式四项变量；针对经产妇的模型则纳入孕前体重指数（Body Mass Index, BMI）、流产/自然流产史、妊娠期SUI及分娩方式。初产妇与经产妇模型的预测概率与实际观察概率均具有良好的校准度，霍斯默-莱梅肖检验（Hosmer-Lemeshow test）结果显示：初产妇组p=0.390，经产妇组p=0.364。经bootstrap重抽样校正后的偏倚校正C统计量：初产妇模型为0.763（95%置信区间[CI]：0.693~0.833），经产妇模型为0.783（95%置信区间[CI]：0.726~0.841）。结论：本研究成功构建了针对初产妇与经产妇的产后压力性尿失禁（postpartum stress urinary incontinence, SUI）预测模型，该方法可为分娩前后产后SUI的高危人群风险预测提供实用工具。