External validation and comparison of prediction models for postoperative stress urinary incontinence after pelvic organ prolapse surgery in Chinese women: a multicenter cohort study

For many women, the pelvic floor is a silent, sturdy foundation that holds the bladder, uterus, and other organs in their proper place. When this support weakens, organs can slip downward, a condition known as pelvic organ prolapse. This often requires surgical repair to push the organs back up and secure them. However, fixing the prolapse can sometimes uncover a hidden problem: stress urinary incontinence. This occurs when a woman leaks urine during coughing, sneezing, or laughing, a symptom that might have been masked by the prolapse itself. Once the organs are lifted, the leak can become apparent. Doctors face a difficult choice: should they add a procedure to tighten the urethra and prevent leaks during the initial surgery, or wait and see? Adding the extra procedure carries risks, such as difficulty emptying the bladder or infection, so performing it on someone who would never have developed leaks is unnecessary. To help make this decision, researchers have created prediction models—tools that use a patient's age, medical history, and physical exam results to estimate her personal risk of developing these leaks after surgery.

The question of whether these tools work for everyone has remained unanswered. Most existing models were built using data from women in the Netherlands or South Korea, where body types, surgical habits, and healthcare systems differ from those in China. A team of researchers from three medical centers in China set out to test if these foreign tools could accurately predict outcomes for Chinese women. They gathered the medical records of 673 women who had undergone surgery for moderate to severe pelvic organ prolapse between 2020 and 2023. These women came from hospitals in Tianjin and Jinhua, representing a real-world mix of patients and surgical approaches. The researchers focused on a specific outcome: did the woman develop bothersome urine leaks within one year of her surgery, or did she need treatment for them? By comparing what the models predicted against what actually happened, the team could see if the tools held up in a new population.

The study revealed that while the prediction models were not perfect, they were far from useless. Out of the 673 women in the study, 151, or about 22 percent, developed bothersome leaks or required treatment within a year. When the researchers applied the Dutch model to this group, it performed reasonably well, providing a generally accurate estimate of the overall risk, though the relationship between predicted and observed outcomes was not perfect. It also showed a moderate ability to distinguish between women who would develop leaks and those who would not. The Korean model, by contrast, struggled more. It tended to overestimate the risk, predicting that more women would have leaks than actually did, and it was less effective at separating high-risk patients from low-risk ones. In simple terms, the Dutch model was closer to the truth, while the Korean model sounded the alarm too often.

The researchers found that the differences in performance likely stem from the unique characteristics of the patients and the surgeries involved. The women in the Chinese cohort were slightly older on average and had different body measurements than the women used to build the original models. Furthermore, the types of surgery performed in China were more varied, including a wide range of techniques from native tissue repairs to the use of mesh, whereas the original models were based on more specific surgical styles. These variations in patient background and medical practice mean that a tool designed for one group does not always translate perfectly to another. The study did not find a new, perfect model for Chinese women, but it did identify that the existing Dutch model is the better of the two options currently available for this population.

Ultimately, the study suggests that these prediction tools can offer valuable guidance, but they must be used with caution. The Dutch model appears to be a helpful resource for doctors and patients in China to discuss the risks and benefits of adding a leak-prevention procedure during prolapse surgery. It can support shared decision-making by providing a personalized risk estimate, helping to balance the potential benefit of preventing leaks against the risks of unnecessary surgery. However, the researchers emphasize that these models are not a replacement for clinical judgment. Before these tools can be widely adopted in everyday practice, they may need to be fine-tuned or recalibrated to fit the specific realities of Chinese patients and surgeons. The work highlights that medical tools are not one-size-fits-all; they must be tested and adapted to ensure they serve the people who need them most.