Can Machine Learning Revolutionize Post-Retrograde Intrarenal Surgery Urosepsis Prediction? A Single-Center Study

Background: One of the main surgical methods for upper urinary stones is retrograde intrarenal surgery (RIRS). Urosepsis is a serious complication of RIRS that threatens patients and confronts clinicians. To construct a valid predictive model for post-RIRS urosepsis, a dataset including demographic and pre-operative factors from 260 patients who underwent RIRS was used. Objective: The aim of this research was to create a machine learning (ML) model as a novel solution to predict high-risk patient populations for urosepsis after retrograde intrarenal surgery (RIRS). Method: This retrospective analysis involved 260 patients who were treated with retrograde intrarenal surgery (RIRS) without pre-stenting at Pakistan Kidney and Liver Institute & Research Center from September 2018 to August 2024. Demographic, clinical, and preoperative data were retrieved to construct a predictive model for post-RIRS urosepsis. Supervised machine learning algorithms, i.e., Support Vector Machine, Gaussian Naïve Bayes, Logistic Regression, Decision Tree, and k-nearest Neighbors, were utilized. Model performance was assessed by accuracy, precision, recall, and Area Under the Receiver Operating Characteristic Curve. Results: The machine learning models were able to predict post-RIRS urosepsis based on preoperative demographic and clinical features. Of the algorithms used, Support Vector Machine (SVM), Logistic Regression, and k-Nearest Neighbors (KNN) classifiers performed best in terms of predictive accuracy, and SVM had the best overall accuracy. The findings prove that ML-based methods are capable of predicting high-risk patients before surgery effectively. Conclusion: This algorithm encompasses the potential to detect and prevent the development of urosepsis in RIRS patients and creating proper care plans through machine learning models.