Integrating Pre- and Postoperative Systemic Inflammatory Markers for Acute Kidney Injury Prediction Following Radical Cystectomy: A Multi-Center Retrospective Study.

Purpose: The present study aimed to investigate the association of perioperative dynamic changes of systemic inflammation markers with AKI after radical cystectomy and their predictive value through machine learning algorithms.

Patients and methods: Patients undergoing radical cystectomy with urinary diversion for bladder cancer from 2013 to 2022 at three university-affiliated tertiary hospitals were gathered. Perioperative dynamic changes of systemic inflammatory markers were calculated based on peripheral blood cell counts from pre- and post-operative values and categorized using restricted cubic splines (RCS). The number of positive changes in these markers was recorded as the perioperative inflammation index. Multivariable logistic regression was utilized to identify risk factors for AKI after radical cystectomy. AKI prediction models were constructed through various supervised machine learning algorithms and evaluated by the area under the receiver operating characteristic curve (AUROC).

Results: 727 patients were finally enrolled in the study, with 151 (20.8%) patients experiencing AKI following radical cystectomy. Postoperative hemoglobin (p = 0.003; OR, 0.977; 95% CI, 0.962-0.992), albumin level (p = 0.007; OR, 0.906; 95% CI, 0.843-0.974), intraoperative fluid infusion rate (p < 0.001; OR, 0.769; 95% CI, 0.665-0.890) and the perioperative inflammation index (p < 0.001; OR, 1.507; 95% CI, 1.209-1.877) were identified as independent risk factors with predictive value for AKI following radical cystectomy with urinary diversion. Among various machine learning models, XGBoost performed best (AUROC: 0.801; 95% CI: 0.735-0.867) in AKI prediction.

Conclusion: The association between perioperative dynamic changes of inflammatory markers and AKI after radical cystectomy reinforced the necessity of perioperative inflammatory evaluation. AKI predictive models, integrating perioperative metrics, enable early identification and optimize perioperative management for AKI prevention.