Development of a machine learning model for predicting renal damage in children with closed spinal dysraphism

Area of Science:

• Pediatric Nephrology
• Computational Medicine
• Urology

Background:

• Renal damage in closed spinal dysraphism (CSD) is often caused by neurogenic bladder dysfunction.
• This damage increases the risk of chronic kidney disease and impacts long-term outcomes.
• Early identification of at-risk patients is crucial for timely interventions and improved bladder management.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting renal damage in pediatric CSD patients.
• To identify key predictors of renal damage in this population.

Main Methods:

• Retrospective analysis of 110 children with CSD.
• Development and comparison of four machine learning models: logistic regression, support vector machine, decision tree, and extreme gradient boosting (XGBoost).
• Evaluation of model performance using AUC, calibration curves, and decision curve analysis; interpretation using SHAP and LIME.

Main Results:

• The XGBoost model demonstrated superior predictive performance with an AUC of 0.957.
• SHAP analysis identified abnormal radiological lower urinary tract findings, female sex, and high-grade vesicoureteral reflux as the most significant predictors.
• The model effectively predicted renal damage in children with CSD.

Conclusions:

• An accurate machine learning model using the XGBoost algorithm was successfully developed to predict renal damage in children with CSD.
• This predictive model holds potential for early risk stratification and intervention to preserve renal function.