Artificial Intelligence ChatGPT4-Generated Statistical Analysis Of Managing Stress Urinary Incontinence in Female Athletes

Introduction and Hypothesis: Stress urinary incontinence (SUI) poses a significant challenge for elite female athletes (EFAs), particularly those with severe postpartum SUI. This study aims to identify predictive factors for a successful return to sports following pelvic floor muscle training (PFMT) in these athletes. We hypothesize that utilizing the artificial intelligence chat model GPT-4 will enable the development of a predictive model for EFAs, facilitating their return to sports. Methods: A retrospective cohort study was conducted with 153 EFAs experiencing postpartum SUI. Data on PFMT frequency, treatment methods, return-to-sport rates at 1 year, and 1-Hr Pad test results at three months were collected. Statistical analyses, including Pearson's univariate correlation, logistic multivariate regression, ROC analysis, and random forest model, were performed utilizing GPT-4. Results: At three months, 26.8% showed improvement in SUI. By one year, 28.1% successfully returned to sports. The predicted logit transformation equation for returning to competition (logit(p)) included serum total testosterone (a), PFMT frequency per week (b), 1-Hr Pad test at 3 months (c), and VEL+UEL therapy (d): -126 - 0.07276a + 25.98b - 1.947c - 25.32d, with a logit(p) cutoff at 0.5. ROC analysis determined optimal cutoff values. The random forest model identified these four variables as influential factors. Conclusion: PFMT frequency and VEL+UEL treatment play a crucial role in facilitating the successful return of EFAs with severe SUI to sports. The utilization of the AI chat model GPT-4 enhances the development of a predictive model for EFAs, providing a better understanding of factors influencing return-to-sport outcomes.