Predicting complications after laparoscopic surgery for ureteropelvic junction obstruction using machine learning...

Postoperative complications in patients with ureteropelvic junction obstruction (UPJO) negatively impact surgical outcomes and may necessitate redo surgery. We aimed to predict the occurrence of postoperative complications in these patients using machine learning algorithms.

Data of UPJO patients admitted to our hospital for surgical treatment from May 2014 to May 2023 were retrospectively analyzed. Risk factors were screened using multivariate logistic and Lasso regression. Logistic regression (LR), k-nearest neighbours (KNN), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), Extreme Gradient Boosting (XGB) and Neural Network (NN) were used to create a prediction model.

526 patients were included, with 97 complications (61 urinary tract infections [UTI] and 36 recurrences). Risk factors for postoperative complications of pyeloplasty were preoperative UTI (Pre-UTI), calculus, renal cortical thickness (RCT), collecting system, time of removal of DJ, removal of drainage, and white blood cell count (WBC). Factors associated with post-UTI were p-UTI, RCT, collecting system, time of removal of DJ, and WBC. Factors influencing postoperative recurrence were p-UTI, calculus, RCT, and drainage removal. Finally, LR was selected to develop the clinical prediction model for postoperative complications, UTIs, and recurrence (area under the curve: 0.929, 0.941, and 0.894, respectively). The present study is the first predictive model on total complications, UTI and recurrence after pyeloplasty and demonstrated strong predictive results. However, there are some limitations; this is a single-center study, and the model has not undergone external validation, which may affect the generalizability of our findings.

UPJO postoperative complications, UTI, and recurrence can be predicted prior to surgery by machine learning.