OR1-3 | A Multicenter Predictive Model for Clinically Meaningful Adverse Events in Pediatric Patients Undergoing Transcatheter Pulmonary Vein Interventions

Pulmonary vein stenosis (PVS) in pediatric patients, a condition requiring frequent transcatheter palliative interventions with high morbidity and mortality, necessitates development of a multicenter predictive model for clinically meaningful adverse events (CMAE). Patient and procedural data were collected for Congenital Cardiac Catheterization Project on Outcomes (C3PO) registry cases with PVS interventions aged ≤18 years from 1/1/19 - 12/31/22. A multivariable logistic regression model predicting the primary outcome CMAE (severity levels 3bc/4/5) was built using a random sample of 75% of cases and validated with the remaining 25%. Model discrimination was assessed using the c-statistic and calibration using the Hosmer-Lemeshow test. The derivation dataset of 716 patients had an overall CMAE rate of 7.8%. Patient and procedural characteristics considered for model inclusion are shown in Table 1. The final multivariable model included male sex (odds ratio [OR] 1.74), genetic syndrome (OR 1.99), admission source (OR for medical unit 2.28; OR for step-down unit or ICU 2.07), case duration (OR for 120-209 minutes 2.05, OR for ≥210 minutes 3.18), and hemodynamic vulnerability score (OR for 1 point = 2.05, 2 points = 4.41, ≥3 points = 3.44). C-statistic 0.75, Hosmer-Lemeshow test p = 0.57. Important predictive features include hemodynamic vulnerability score, case duration as a marker for procedural complexity and admission source. Improved understanding of risk and future use of predictive modeling may allow for development of risk mitigation strategies to improve outcomes.