Henry P. Foote, Minghui Sun, Benjamin Alan Goldstein, Kevin D. Hill, Rachel G. Greenberg, Samuel J. Gentle, Kanecia O. Zimmerman, Rishikesan Kameleswaran, Veeral N. Tolia, Matthew M. Laughon, Wesley Jackson, Christoph P. Hornik
Duke University Medical Center and Duke University. Yale University School of Medicine. Baylor University Medical Center. University of North Carolina.
United States
Journal of Perinatology
J Perinatol 2026;
DOI: 10.1038/s41372-026-02576-2
Abstract
Objective: Develop and validate predictive models for pulmonary hypertension (PH) in high-risk infants.
Study design: We trained logistic regression (LR) and long short-term memory (LSTM) models using a multicenter cohort study of infants 22-28 weeks gestational age discharged from neonatal intensive care units from 2008 to 2020, at two timepoints: 33 weeks post-menstrual age (PMA) for infants receiving mechanical ventilation at that time, and 36 weeks PMA for infants receiving any respiratory support at that time.
Results: At 33 weeks PMA (N = 2849), top LR model predictors were current fraction of inspired oxygen and birth weight. At 36 weeks (N = 20,173), top LR model predictors were current respiratory support and birth weight. Both LR and LSTM models had strong performance in the temporal validation cohort (infants discharged 2021-2022) for both timepoints.
Conclusion: Using available clinical variables, we developed and validated predictive models that may identify infants most at risk for PH at two timepoints.
Category
Class III. Pulmonary Hypertension Associated with Lung Disease
Mechanical and Computer Models of Pulmonary Vascular Disease and Therapy
Age Focus: Pediatric Pulmonary Vascular Disease
Fresh or Filed Publication: Fresh (PHresh). Less than 1-2 years since publication
Article Access
Free PDF File or Full Text Article Available Through PubMed or DOI: No