057 - Predictors of Adverse Outcomes in Preterm Infants With Pulmonary Hypertension

Publication Number: 57.6333

Claire E. Johnson, University of Alabama School of Medicine, Homewood, AL, United States; Vivek Shukla, University of Alabama at Birmingham, Birmingham, AL, United States; Amr Hammouda, University of Alabama School of Medicine, Birmingham, AL, United States; Avinash Singh, Children's of Alabama, Birmingham, AL, United States; Matthew Clark, University of Alabama-Birmingham, Trussville, AL, United States; Waldemar Carlo, UAB School of Medicine, Birmingham, AL, United States; Namasivayam Ambalavanan, University of Alabama School of Medicine, Birmingham, AL, United States; Samuel Gentle, Yale School of Medicine, New Haven, CT, United States

Background: Preterm infants who develop BPD-associated pulmonary hypertension (BPD-PH) have up to a 5-fold increased risk for death compared to infants with BPD alone, and it is therefore necessary to define risk factors and predict which patients with BPD-PH will develop severe disease.

Objective: We hypothesized that clinical, demographic, and echocardiographic characteristics available at the time of diagnosis could predict adverse BPD-PH-related outcomes, including BPD-PH persistence at discharge and BPD-PH associated mortality, and exposure to pharmacologic treatment(s).

Design/Methods: This was an observational single center cohort study (2017 and 2024) performed at the University of Alabama at Birmingham in preterm infants born at < 29 weeks’ gestation. Infants included in the cohort were diagnosed with PH based on echocardiographic findings, including tricuspid regurgitation (TR) jet velocity >2.5 m/s, bidirectional flow at the patent ductus arteriosus or foramen ovale, or an eccentricity index >1.1. Adverse BPD-PH-related outcomes included PH persistence at discharge, PH associated mortality, and pharmacologically treated PH. Machine learning models considered demographic, clinical, and echocardiographic characteristics at the time of diagnosis. Models included were Decision Tree, XGBoost, AdaBoost, LightGBM, and Histogram Boosting.

Results: Among the 139 patients diagnosed with BPD-PH, 77 had PH persistence at discharge, 60 were pharmacologically treated, and 27 had PH associated mortality. Infants with PH associated mortality had lower birth weight (median 576; IQR: 145 vs median 624; IQR: 187; p=0.03), were more commonly invasively ventilated at the time of diagnosis (63% vs. 28%; p< 0.001), and had a higher eccentricity index (median 1.3; IQR 0.9 vs median 1.1; IQR 0.3,Table 1). Overall only the models for pharmacologically treated PH and PH associated mortality had a clinically useful predictive performance (AUC >0.70, Table 2). The top predictors in the best performing models for PH associated mortality was respiratory support and pharmacologically treated PH was iNO (Table 3).

Conclusion(s): Variables available at the time of BPD-PH diagnosis were able to predict PH associated mortality and pharmacologically treated PH in preterm infants

Table 1: Baseline characteristics
Table 1-Baseline Characteristics.pdf

Table 2: Model performance comparisons
Table 2-Model performance comparisons.pdf

Table 3: Top Predictors
Table 3-Top Predictors.pdf