Efficacy of left internal spermatic vein reflux reconstruction under microscope left spermatic vein high ligation for treatment of degree III left spermatic vein Significance of influencing factors Ranking and Construction and validation of predictive model

Objective To analyze the importance ranking of influencing factors of postoperative complications in patients with ureteral calculi using Logistic regression and random forest algorithm, and to construct and verify the prediction model. Methods Clinical information of 254 patients with ureteral calculi who received surgical treatment in our hospital from January 2021 to December 2023 was collected as variables, including age, gender, calculi diameter, calculi location, calculi type, preoperative urine routine, etc. Patients were divided into complication group (n=52) and no-complication group (n=202) according to whether there was any postoperative complication. Univariate and multivariate analyses using Logistic regression were performed to identify independent risk factors for postoperative complications. The random forest algorithm was further used to construct the prediction model, and the performance of the model was verified by receiver operating characteristic curve (ROC) analysis, calibration curve evaluation, and decision curve analysis (DCA). Results A total of 254 patients were included, and 52 cases (20.47%) had postoperative complications. Logistic regression analysis showed that hypertension, diabetes, neurogenic bladder, preoperative urinary tract infection, stone diameter, operation time and intraoperative perfusion pressure were the independent risk factors for postoperative complications in patients with ureteral calculi (P<0.05). The order of importance of variables obtained from the random forest model was operation time, stone diameter, perfusion pressure during operation, presence of urinary tract infection before operation, concomitant neurogenic bladder, concomitant diabetes mellitus and concomitant hypertension. Nomograms model predicts AUC of 0.801 (95% CI: 0.719-0.883) for postoperative complications of ureteral calculi. The predicted curve lines of the model group and the verification group were basically fitted with the standard curves. Analysis of the decision curve shows that when the probability threshold of Nomograms model for predicting postoperative complications in patients with ureteral calculi is 0.1–0.9, the net benefit rate for patients is large. Conclusion Logistic regression and random forest algorithm can effectively analyze the influencing factors of postoperative complications in patients with ureteral calculi and construct accurate prediction model. The model can provide a theoretical reference for the prevention and treatment of postoperative complications and optimize the individualized treatment plan.