关键词: 尿失禁, 压力性；女性；Meta分析；Rothman-Keller模型；预测

Abstract: Background　As the most common type of female urinary incontinence，stress urinary incontinence（SUI） has an increasing incidence in adult women with the process of aging and lifestyle changes in recent years. Although it is not life-threatening，it seriously affects the performance of work and physical and mental health. Objective　To develop a Rothman-Keller-type model for actively screening for SUI risk in adult women，so as to effectively identify and timely intervene with the high-risk groups. Methods　Observational studies about SUI in adult women were searched from databases of PubMed，EMBase，SinoMed，CNKI and Wanfang Data from January 2005 to December 2020. Review Manager 5.3 software was used for meta-analysis. The establishing principles of the Rothman-Keller model were used for creating a Rothman-Keller-type model for predicting the risk of SUI in adult women，and the threshold of different risks was determined using simulation data sets. The predictive accuracy of the model was verified by using Logistic regression model. Results　Finally，18 articles were included.The 12 risk factors obtained from analysis were：age（≥50 years），BMI（≥24 kg/m2），menopause，urinary system infection，vaginal delivery，and pelvic prolapse，history of pelvic surgery，number of deliveries（≥3 times），constipation，chronic respiratory disease，hypertension，and diabetes. We generated 20 000 simulated individual data based on the risk factors using binomial distribution function，and incorporated them into the Rothman-Keller-type model to plot a curve predicting the SUI risk in adult women，and chose 0.285 5 and 0.429 4 as the threshold for dividing low and intermediate risks，and intermediate and high risks，respectively，according to the curve changes. We incorporated the simulated individual data into a Logistic regression model to plot a curve predicting the SUI risk in adult women，and used 0.261 2 and 0.547 4 as the threshold for dividing low and intermediate risks，and intermediate and high risks，respectively，according to changes in the curve. Validation results of risk prediction for a 60-year-old female：she was predicted to have a high risk for SUI〔P=0.485 6（>0.429 4）〕 by the Rothman-Keller-type model，and was classified into high risk group for SUI 〔P=0.878 6（>0.547 4）〕 according to the prediction using the Logistic regression model，indicating that the predictive results of the two models were consistent. Conclusion　Our model has proved to have strong accuracy and predictability. Our experience may provide ideas for the development of risk prediction models for other chronic diseases.

Key words: Urinary incontinence, stress；Femininity；Meta-analysis；Rothman-Keller model；Forecasting