Derivation and Validation of Clinical Prediction Model of Surgical Risk and Renal Scan Obstruction in Children with Isolated Unilateral Antenatal Hydronephrosis

Abstract

Introduction: There is no single agreed-upon tool to forecast hydronephrosis obstruction and surgical risk in the literature. Previous scoring systems have relied on renal scan parameters for this task, which are not performed or read consistently across hospitals. We aim to determine the utility of a hydronephrosis obstruction early scoring system in identifying severe UPJO-like hydronephrosis suggestive of surgery. Methods: We reviewed a prenatal hydronephrosis database (2008-23) to identify patients with unilateral UPJO-like hydronephrosis. Patients with vesicoureteral reflux (VUR), megaureter, or other anomalies were excluded. We collected data on the first ultrasound and surgical outcome for the Society for Fetal Urology (SFU) grade, anteroposterior pelvic diameter (APD), and percent difference in renal length – the renal length index (RLI). I undertook a multivariable logistic regression to determine the predicted value of these sonographic measurements. Model development and internal validation were performed, and model performance (calibration and discrimination) was reported. Results: Of the 465 UPJO-like hydronephrosis patients, 76.1% (354) were male and had a mean age at baseline of 3.8 ± 3.8 months. 34.8% (162) had an obstructive renal scan and 27.7% (129) underwent pyeloplasty. The developed model to predict surgery using the aforementioned variables with excellent model performance (AUC 0.92). A simplified model using SFU 3-4, APD > 13mm and RLI >12% also predicted surgery with great model performance (AUC 0.89) and improved clinical user-friendliness. Conclusion: Our clinical predictive model can accurately predict surgery in both derivation and validation cohorts of patients with antenatal UPJO-like hydronephrosis. This non-invasive tool can greatly aid in patient counselling.