Development and validation of a prediction model for the risk of postoperative renal dysfunction in patients with renal cancer based on machine learning

Objective To construct a risk prediction model for postoperative renal dysfunction based on preoperative routine clinical data using machine learning techniques, providing decision support for the early identification of high-risk patients.

Methods A total of 413 postoperative renal dysfunction in renal cell carcinoma (RCC) patients who underwent partial or radical nephrectomy at the First Affiliated Hospital of Guangxi Medical University were included. Key variables were selected from 79 initial features using LASSO regression. Nine machine learning algorithms, including XGBoost, random forest, and logistic regression, were used to construct the prediction model. The model's performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA). The SHAP method was employed to analyze the contribution of predictive factors.

Results The logistic regression model performed best in the validation cohort, with an AUC of 0.798 (95% CI: 0.646-0.948). SHAP analysis identified endogenous creatinine clearance, age, and the maximum tumor diameter as key predictive factors. DCA indicated that the model provides significant clinical net benefit, and the calibration curve suggested good predictive calibration.

Conclusion The logistic regression-based prediction model can effectively identify high-risk patients for postoperative renal dysfunction in RCC. Its simplicity offers significant advantages, and it has potential for clinical application.