Nonlinear logistic regression model for outcomes after endourologic procedures: a novel predictor

Area of Science:

• Urology
• Medical Informatics
• Biostatistics

Background:

• Endourologic interventions are common for treating kidney stones.
• Predicting outcomes after these procedures is crucial for patient management.
• Existing models may not fully capture the complexity of treatment outcomes.

Purpose of the Study:

• To design a practical nonlinear logistic regression model for predicting outcomes after endourologic interventions.
• To assess the model's performance in predicting stone-free status and the need for secondary procedures.

Main Methods:

• A nonlinear logistic regression model was developed and cross-validated using data from 382 renal units.
• Input variables and outcome data from single-institution endourologic treatments were utilized.
• Model performance was evaluated using sensitivity, specificity, PPV, NPV, and ROC AUC.

Main Results:

• The model predicted stone-free status with 69.6% accuracy (ROC AUC 0.749).
• It predicted the need for secondary procedures with high accuracy (ROC AUC 0.863), comparable to traditional models.
• The model demonstrated proof-of-concept for predicting outcomes after shockwave lithotripsy, ureteroscopic lithotripsy, and percutaneous nephrolithotomy.

Conclusions:

• Nonlinear logistic regression can effectively predict key clinical outcomes in endourology.
• The developed model shows potential for future optimization with larger, multi-institutional datasets.
• This approach could lead to the development of predictive nomograms for endourologic procedures.