Related Experiment Videos
Machine learning-based prediction of Clavien-Dindo ≥ II complications after ureteroscopy
Miroslav Stojadinović1, Slobodan Janković2
1Faculty of Medical Sciences, University of Kragujevac, Kragujevac, Serbia. midinac@gmail.com.
International Urology and Nephrology
|August 1, 2026
Summary
This study developed a prediction model for complications after ureteroscopy, identifying operative time and impacted stones as key predictors. The model aids in personalized risk assessment for patients undergoing ureteroscopic lithotripsy.
Area of Science:
- Urology
- Medical Informatics
- Predictive Analytics
Background:
- Clavien-Dindo (CD) grade II complications occur in 13-17% of ureteroscopy cases, often requiring medical intervention for infections or bleeding.
- Predicting these complications is crucial for patient management and preoperative counseling.
Purpose of the Study:
- To develop and internally validate a prediction model for CD grade ≥II complications following ureteroscopy.
- The model utilizes routinely available perioperative predictors for lean application.
Main Methods:
- Retrospective analysis of 389 ureteroscopic lithotripsy patients (2010-2014).
- Models (binomial generalized linear model with Elastic Net, gradient boosting model) were trained and tested on an 80:20 split.
- Performance evaluated using area under the receiver operating characteristic curve (AUC), calibration, accuracy, and decision curve analysis.
Main Results:
- 8.5% of patients experienced clinically significant complications.
- Both developed models showed excellent discrimination (GLM AUC 0.893, GBM AUC 0.857).
- Operative time and impacted stone status were identified as significant predictors of complications.
Conclusions:
- The developed prediction model effectively stratifies individualized risk for Clavien-Dindo grade ≥II complications after ureteroscopy.
- Integrating perioperative variables supports preoperative counseling and clinical decision-making.
