Machine learning-based prediction of vesicoureteral reflux outcomes in infants under antibiotic prophylaxis.
Journal:
Scientific reports
PMID:
40121274
Abstract
We aimed to investigate the independent outcome predictors of continuous antibiotic prophylaxis (CAP) in vesicoureteral reflux, train a model to predict the outcome, and evaluate which infants should be referred for endoscopic vesicoureteral reflux correction in their first visits. A total of 225 infants ≤ 2 years of age with a diagnosis of vesicoureteral reflux between 2009 and 2022 were recruited; 115 patients from a pediatric nephrology clinic received CAP, and 110 patients from a pediatric urology department underwent endoscopic injection of dextranomer/hyaluronic acid copolymer. In the multivariable analysis, only renal scarring and bladder dysfunction were significantly associated with post-treatment febrile urinary tract infections and/or renal scarring and vesicoureteral reflux persistence, respectively, in children who received CAP. The machine learning modeling showed that for both febrile urinary tract infections and/or renal scarring and vesicoureteral reflux persistence, the random forest was the best fit. On the other hand, we observed that the success rates of endoscopic injection among the patients with renal scarring and bladder dysfunction were acceptable. In conclusion, renal scarring and bladder dysfunction were predictors of vesicoureteral reflux outcomes when the infant was receiving CAP. Therefore, referring these patients to a urologist is advised during their first visits as they benefit from endoscopic injection.