Machine Learning and Latent Profile Analysis Predict the Health-Related Quality of Life of Bladder Cancer Patients with Abdominal Wall Stomas

Abstract Abdominal wall stomas in bladder cancer patients can severely impact physical, psychological, and social adjustment. However, individual differences in patients' self-reported outcomes and their influencing factors require further investigation. This multicenter, cross-sectional study aimed to assess the health status of patients with bladder cancer and abdominal wall stomas, identify heterogeneous characteristics among individuals, and screen for risk factors affecting their quality of life. We evaluated 598 patients who underwent radical cystectomy with an abdominal wall stoma between September 2020 and September 2023 in the urology departments of five tertiary hospitals across Henan and Jiangxi Provinces, China. Data were collected via a self-designed general profile questionnaire and the Chinese Patient Self-Reported Outcome Scale for Bladder Cancer with Abdominal Wall Stoma (PROS-BCAS). Latent profile analysis identified three patient subgroups: "Balanced high-quality type" (n=339, 56.7%), "Unbalanced type" (n=148, 24.7%), and "Low-level type" (n=111, 18.6%). Significant differences were observed in the scores of the four PROS-BCAS domains among these categories (P<0.05). The factors associated with the "Low-level type" group included age (45--60 years), low income (monthly per capita <2,000 CNY), self-payment for healthcare, total lack of self-care ability, and the presence of comorbidities. Among the machine learning models evaluated, the XGBoost model statistically outperformed others (e.g., random forest, logistic regression) in predicting patient prognosis, achieving an AUC of 0.89 and an average F1-score of 0.89. The model identified key prognostic factors in descending order of importance: per capita monthly household income, age, medical payment method, status of stoma care, and complications. These findings highlight the significant individual heterogeneity in self-reported outcomes among bladder cancer patients with abdominal wall stomas across Henan, Jiangxi Provinces, and Beijing, China. Clinically, this study provides actionable insights for identifying high-risk subgroups and tailoring interventions. For example, caregivers should prioritize vulnerable populations by addressing economic barriers, optimizing stoma care, and providing targeted psychological and social support. These measures can enhance patient quality of life and inform precision care strategies for bladder cancer patients with stomas.