The ductus arteriosus is a normal feature of fetal circulation, a short muscular vessel that shunts blood away from the lungs while the baby is still in the womb and oxygen comes from the placenta. In most newborns, the vessel closes within hours or days of birth. In babies born before 32 weeks of gestation, however, closure often fails or is delayed. The resulting patent ductus arteriosus, or PDA, creates a left-to-right shunt: oxygenated blood leaving the heart is diverted back toward the lungs, increasing pulmonary blood flow and pressure. When that shunt becomes hemodynamically significant, meaning it is large enough to disturb circulatory function, the flooded, immature lung capillaries are thought to become vulnerable to rupture, particularly in lungs already injured by prematurity and mechanical ventilation.

The research team, led by Li Yang and colleagues in the Department of Neonatology at Guangyuan Central Hospital in Sichuan, China, set out to quantify this relationship. They reviewed the records of 482 very preterm infants, all born at less than 32 weeks of gestation, who were admitted to their neonatal intensive care unit between January 2019 and October 2025. Because the study was retrospective, the investigators worked backward from outcomes that had already occurred, using univariate and multivariate logistic regression to separate factors genuinely associated with pulmonary hemorrhage from those that merely appeared linked because of confounding. For infants diagnosed with a patent ductus arteriosus, the team calculated the Patent Ductus Arteriosus Severity Score, a composite measure that grades the clinical and hemodynamic weight of the defect.

The headline finding is stark. Of the 482 infants in the cohort, 93, or 19.3 percent, developed neonatal pulmonary hemorrhage. Among those who bled, 42 died before discharge, a post-hemorrhage mortality rate of 45.2 percent. Those numbers underline why the condition commands such dread in neonatology: nearly one in five very preterm infants in this cohort experienced a lung hemorrhage, and nearly half of the affected babies did not survive it. Preventing even a fraction of these events, or identifying them earlier, could translate into a meaningful survival gain for this population.

When the researchers ran their multivariate analysis across the entire cohort, five factors remained independently associated with pulmonary hemorrhage: lower gestational age, lower birth weight, a lower five-minute Apgar score, premature rupture of membranes, and hemodynamically significant PDA. Each of these tells a clinically coherent story. Earlier gestational age and lower birth weight reflect pulmonary immaturity and fragile capillary networks. A depressed five-minute Apgar score signals birth depression and possible hypoxic stress on vascular integrity. Premature rupture of membranes is associated with intrauterine inflammation and infection, which can precondition the fetal lung for injury. And hemodynamically significant PDA supplies the mechanical mechanism, the excessive pulmonary blood flow, that pushes those weakened capillaries over the edge.

To make these findings more usable at the bedside, the team built nomograms, graphical scoring tools that allow clinicians to total up a patient’s individual risk factors and read off an estimated probability of the outcome. They internally validated these tools using the bootstrap method, a resampling technique that estimates how much a model’s apparent performance is inflated by overfitting to the original dataset. In the subgroup of infants with PDA, the nomogram for predicting pulmonary hemorrhage achieved an area under the receiver operating characteristic curve, or AUC, of 0.858, with a 95 percent confidence interval of 0.812 to 0.904. After bootstrap correction, the value settled at 0.837, still comfortably in the range considered good discrimination for a clinical prediction tool.

One of the more contemporary aspects of the study is its use of SHAP analysis, short for SHapley additive explanations, a machine learning interpretation method borrowed from game theory that quantifies how much each individual predictor contributes to a model’s output. Applied here, SHAP revealed that gestational age contributed most to the risk assessment of pulmonary hemorrhage, followed by hemodynamically significant PDA and then birth weight. In practical terms, this means the single strongest signal of danger remains how early the baby was born, but the severity of the ductal shunt carries weight nearly comparable to birth weight. The approach gives clinicians an intuitive visualization of why the model flags a particular infant, addressing a common criticism that statistical prediction tools operate as unexplained black boxes.

The second half of the analysis focused on the sickest subgroup: the 81 infants who had both a patent ductus arteriosus and a pulmonary hemorrhage. Here the question shifted from who bleeds to who dies after bleeding. In multivariate analysis, a higher Patent Ductus Arteriosus Severity Score was independently associated with in-hospital mortality, with an odds ratio of 3.614 and a 95 percent confidence interval of 1.918 to 6.812. In plain language, each step up in severity score roughly tripled or more the odds of dying, a strong and statistically robust association whose confidence interval lies well above the null. This suggests the score, which aggregates indicators of ductal hemodynamic burden, may function as an early warning marker for the babies least likely to survive a hemorrhagic event.

The clinical implications are considerable. If hemodynamically significant PDA is independently linked to pulmonary hemorrhage, then the intensity of monitoring and treatment of the ductus in very preterm infants, whether through fluid restriction, pharmacologic closure, or other strategies, may matter not only for classic ductal outcomes but for lung bleeding itself. Meanwhile, the severity score’s association with mortality offers a way to stratify infants with PDA who do hemorrhage, helping clinicians and families calibrate expectations and direct intensive resources toward those at highest risk. The nomograms, once validated, could become routine bedside calculators in neonatal units caring for extremely premature babies.

The authors themselves are careful about the limits of what they can claim. As a single-center retrospective cohort, the study cannot establish that treating the ductus prevents hemorrhage; association is not causation, and unmeasured variables may still influence both conditions. The internal bootstrap validation, while methodologically sound, does not substitute for external validation in larger, independent cohorts, which the researchers explicitly identify as the necessary next step before the tools enter routine practice. Still, by combining a large consecutive cohort, adjusted regression modeling, machine learning interpretability, and validated risk scores, the study offers one of the clearest quantitative portraits to date of how a simple failure of fetal circulation to close can help push a preterm lung into its most dangerous complication, and how a structured score may tell clinicians, early, which of these babies are in the gravest danger.

Subject of Research: Association between patent ductus arteriosus and neonatal pulmonary hemorrhage and mortality in very preterm infants

Article Title: Association of patent ductus arteriosus and severity score with neonatal pulmonary hemorrhage and mortality in very preterm infants: a retrospective cohort study

Article References: Association of patent ductus arteriosus and severity score with neonatal pulmonary hemorrhage and mortality in very preterm infants: a retrospective cohort study. (n.d.). https://doi.org/10.1186/s12887-026-07663-z

Image Credits: AI Generated

DOI: 10.1186/s12887-026-07663-z

Keywords: patent ductus arteriosus, very preterm infants, neonatal pulmonary hemorrhage, PDA severity score, mortality, nomogram, SHAP analysis, logistic regression, neonatology, risk factors, retrospective cohort study, BMC Pediatrics