Wang Qi, Jiang Xiaoxiang, Li Xiaoyan, Que Yanzhen, Lin Chaoqin
Department of Gynecology, Fujian Maternity and Child Health Hospital, 18 Dao-Shan Street, Gu-Lou District, Fuzhou, 350000, PR China.
College of Clinical Medicine for Obstetrics & Gynecology and Pediatrics, Fujian Medical University, Fuzhou, PR China.
Int Urogynecol J. 2025 Feb;36(2):391-401. doi: 10.1007/s00192-024-05983-1. Epub 2024 Nov 25.
Accurate identification of female populations at high risk for urinary incontinence (UI) and early intervention are potentially effective initiatives to reduce the prevalence of UI. We aimed to apply machine-learning techniques to establish, internally validate, and provide interpretable risk assessment tools.
Data from a cross-sectional epidemiological survey of female urinary incontinence conducted in 2022 were used. Sociodemographic and obstetrics-related characteristics, comorbidities, and urinary incontinence questionnaire results were used to develop multiple prediction models. Seventy percent of the individuals in the study cohort were employed in model training, and the remainder were used for internal validation. Model performance was characterized by area under the receiver-operating characteristic curve (AUC) and calibration curves, as well as Brier scores. The best-performing model was finally selected to develop an online prediction tool.
The results showed that bothersome stress urinary incontinence (BSUI) occurred in 9.6% (849 out of 8,830) of parous women. The XGBoost model achieved the best prediction performance (training set: AUC 0.796, 95% confidence interval [CI]: 0.778-0.815, validation set: AUC 0.720, 95% CI: 0.686-0.754). Additionally, the XGBoost model achieved the lowest (best) Brier score among the models, with sensitivity of 0.657, specificity of 0.690, accuracy of 0.688, positive predictive value of 0.231, and negative predictive value of 0.948. Based on this model, the top five risk factors for the development of BSUI among parous women were ranked as follows: body mass index, age, vaginal delivery, constipation, and maximum fetal birth weight. An online calculator was provided for clinical use.
The application of machine-learning algorithms provides an acceptable, though not perfect, prediction of BSUI risk among parous women, requiring further validation and improvement in future research.
准确识别尿失禁（UI）高危女性群体并进行早期干预，可能是降低尿失禁患病率的有效举措。我们旨在应用机器学习技术来建立、内部验证并提供可解释的风险评估工具。
使用了2022年开展的一项女性尿失禁横断面流行病学调查的数据。社会人口统计学和产科相关特征、合并症以及尿失禁问卷结果被用于开发多个预测模型。研究队列中70%的个体用于模型训练，其余个体用于内部验证。模型性能通过受试者操作特征曲线（AUC）下面积、校准曲线以及Brier评分来表征。最终选择表现最佳的模型来开发在线预测工具。
结果显示，经产妇中令人困扰的压力性尿失禁（BSUI）发生率为9.6%（8830例中的849例）。XGBoost模型表现出最佳预测性能（训练集：AUC 0.796，95%置信区间[CI]：0.778 - 0.815；验证集：AUC 0.720，95% CI：0.686 - 0.754）。此外，XGBoost模型在各模型中Brier评分最低（最佳），敏感性为0.657，特异性为0.690，准确性为0.688，阳性预测值为0.231，阴性预测值为0.948。基于该模型，经产妇发生BSUI的前五大风险因素排序如下：体重指数、年龄、阴道分娩、便秘和最大胎儿出生体重。提供了一个在线计算器供临床使用。
机器学习算法的应用为经产妇的BSUI风险提供了一个可接受但并不完美的预测，在未来研究中需要进一步验证和改进。
