Posterior Urethral Valves Outcomes Prediction (PUVOP): a machine learning tool to predict clinically relevant outcomes in boys with posterior urethral valves
Jethro CC Kwong, MD1, Adree Khondker, BHSc2, Jin Kyu Kim, MD1, Michael Chua, MD2, Daniel T. Keefe, MD2, Joana Dos Santos, MD2, Marta Skreta, MSc2, Lauren Erdman, MSc2, Neeta D'Souza, BSc3, Antione F. Selman, MD3, John Weaver, MD3, Gregory Tasian, MD3, Chia Wei Teoh, MD2, Mandy Rickard, NP2, Armando J. Lorenzo, MD2.
1University of Toronto, Toronto, ON, Canada, 2The Hospital for Sick Children, Toronto, ON, Canada, 3Children's Hospital of Philadelphia, Philadelphia, PA, USA.
BACKGROUND: One-third of patients with posterior urethral valve (PUV) are expected to progress to chronic kidney disease (CKD) by adulthood. Due to inevitable myogenic failure, one-quarter will require assistance with bladder emptying with clean-intermittent-catheterization (CIC). Previous studies have shown early risk factors for disease progression in PUV patients, such as baseline kidney function, reflux, or kidney dysplasia; however, they have not been applied collectively to individualize patient care. The objective of this study is to predict progressive decline in renal function and need for additional procedures in patients with PUV using machine learning (ML).
METHODS: Patients diagnosed with PUV with kidney function measurements at our institution between 2000 and 2020 were included. Patients were excluded if there was incomplete data or <1 year follow-up. Pertinent clinical measures were abstracted, including estimated glomerular filtration rate (eGFR) at each visit, initial vesicoureteral reflux grade, and renal dysplasia (abnormal echogenicity, cortical cysts) at presentation. ML models (random survival forests) were developed to predict clinically-relevant outcomes: progression in CKD stage, initiation of renal replacement therapy (RRT), and need for CIC. Model performance was assessed by concordance index (c-index) and the model was externally validated.
RESULTS: A total of 103 patients were included with a median follow-up of 5.7 years. Of these patients, 26 (25%) had CKD progression, 18 (17%) required RRT, and 32 (31%) were prescribed CIC. Additionally, 22 patients were included for external validation from another institution. The ML model predicted CKD progression (c-index = 0.77; external c-index: 0.78), RRT (c-index = 0.95; external c-index: 0.89), and indicated CIC (c-index = 0.70; external c-index: 0.64) and all performed better than Cox proportional-hazards regression. The models have been packaged into a simple easy-to-use tool, available at https://share.streamlit.io/jcckwong/puvop/main/app.py (Figure 1).
CONCLUSIONS: ML-based approaches for predicting clinically-relevant outcomes in PUV are feasible. Further validation is warranted, but this implementable model can act as a decision-making aid for more individualized patient care. 
Back to 2021 Abstracts