Machine learning models to predict systemic inflammatory response syndrome after percutaneous nephrolithotomy.

Area of Science:

• Urology
• Medical Informatics
• Machine Learning

Background:

• Systemic inflammatory response syndrome (SIRS) is a potential complication following percutaneous nephrolithotomy (PCNL).
• Accurate prediction of SIRS is crucial for patient management and surgical decision-making.
• Machine learning offers a promising approach for developing predictive models in healthcare.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting SIRS after PCNL.
• To identify key clinical variables contributing to SIRS prediction.
• To assess the performance of different machine learning algorithms for this task.

Main Methods:

• Retrospective review of clinical data from 337 PCNL patients.
• Development of prediction models using six machine learning algorithms.
• Evaluation of model performance using AUC, accuracy, sensitivity, and specificity on a testing set.

Main Results:

• The Support Vector Machine (SVM) model demonstrated the highest performance (accuracy: 0.868, AUC: 0.942).
• Key predictors identified by the SVM model include prealbumin, preoperative urine culture, SII, and NLR.
• Other significant predictors included staghorn stones, fibrinogen, operation time, and WBC count.

Conclusions:

• Machine learning models can accurately predict SIRS risk post-PCNL.
• These models can assist surgeons in clinical decision-making.
• Prealbumin and inflammatory markers are significant predictors of SIRS after PCNL.