Even with significant progress in neonatal intensive care, extremely premature infants (gestational age <32 weeks) still face a high risk of intraventricular hemorrhage (IVH) (1). This enduring challenge is likely associated with increased survival rates, which, while enhancing overall survival, also continue to pose significant risks to neurodevelopment. IVH is the prevalent form of intracranial hemorrhage in premature infants, affecting 25%–40% of very low birth weight (VLBW) neonates (2, 3). The most severe cases (classified as Volpe grades III–IV) (4) make up roughly 10%–15% of these incidents (2, 3). IVH occurrences in VLBW often take place within the first week, peaking during the first three days (5). The primary factors contributing to IVH involve unstable blood flow during the perinatal period, disrupted brain blood flow regulation, and severe swings in oxygen levels (6, 7). Extremely premature infants have underdeveloped blood vessels in the brain, which are delicate and struggle to maintain stable circulation. This makes them highly susceptible to hemorrhage under conditions such as blood pressure instability, hypoxemia, or hypercapnia (8, 9). Additionally, infections and inflammatory conditions—like maternal chorioamnionitis or sepsis in newborns—worsen the risk of IVH by triggering systemic inflammatory responses and compromising the integrity of the blood-brain barrier (10, 11).

One of the most serious outcomes of IVH is damage to white matter, which can lead to both immediate neurological issues and lasting developmental challenges like cerebral palsy and cognitive impairments (1, 12–14). Therefore, it is crucial to develop methods that can directly evaluate the efficacy of interventions. These approaches will establish a clinical basis for effective IVH prevention strategies, such as optimizing hemodynamic management and controlling infection and inflammatory responses, with the overarching aim of reducing IVH incidence and improving long-term effects in extremely preterm newborns.

To address the unmet need for early, actionable risk stratification in extremely preterm infants, this study developed and validated a novel predictive nomogram for severe IVH during the critical first postnatal week. Our approach uniquely combines time sensitive clinical indicators measurable within 48 h after birth, including initial hemoglobin levels and septic shock status, with dynamic physiological parameters and established prenatal/perinatal risk factors into a clinically implementable tool. This comprehensive modeling strategy enables real time risk assessment during the optimal window for early interventions such as hemodynamic optimization and targeted anti-inflammatory therapies. The nomogram provides probabilistic outputs to support clinical decision making, offering guidance for individualized treatment plans, NICU resource allocation, and informed parental counseling, thereby addressing persistent gaps in current neonatal practice.

In this study, we constructed a nomogram for forecasting the likelihood of critical IVH in exceptionally premature newborns within their initial life week. Using LASSO and logistic regression, we identified six key variables: gestational age, 5-min Apgar score, septic shock, pulmonary hemorrhage, hemoglobin level, and thrombocyte count. Based on these factors, we constructed a model that effectively assesses the risk of severe IVH, demonstrating robust accuracy and strong discriminatory power. Our findings not only corroborate previously reported associations between perinatal risk factors and IVH, but also underscore the synergistic role of hematological parameters and clinical complications in the pathophysiology of IVH.

A substantial body of evidence establishes a strong association between lower gestational age and an increased risk of severe IVH (16–18). In premature infants, IVH primarily occurs due to underdeveloped blood vessels in the brain. The blood-brain barrier, which relies on a complex network of endothelial cells, tight junctions, basement membrane structures, and astrocyte projections, matures progressively throughout gestation—making its stability directly tied to the infant's stage of fetal development. In preterm birth, insufficient glial fibrillary acidic protein around germinal matrix blood vessels weakens vascular integrity. Additionally, decreased fibronectin levels and reduced collagen content result in a poorly organized basement membrane structure. Downregulation of tight junction proteins (e.g., claudin-5) and aquaporin-4 further exacerbates blood-brain barrier's permeability. These structural and molecular deficits lead to increased vascular fragility, impaired barrier function, and an elevated risk of blood extravasation and hemorrhage (17). Moreover, the underdeveloped smooth muscle layer in cerebral vessels of preterm infants weakens the vasoconstrictive response mediated by α1-adrenergic receptors, impairing the capacity to buffer sudden increases in blood pressure. As a result, cerebral vessels are unable to autoregulate blood flow during hypertensive episodes, transmitting excessive perfusion pressure directly to the fragile capillary network, thereby increasing the risk of vessel rupture. Conversely, hypotension may trigger cerebral ischemia (19, 20). Collectively, these findings suggest that gestational age is not only a biological marker of IVH risk but also a critical determinant for guiding individualized monitoring strategies, such as targeted blood pressure management and restrictive ventilation approaches.

Early postnatal clinical factors are key determinants of IVH in extremely preterm infants. The Apgar score is a commonly employed method for evaluating a newborn's health immediately after delivery. Lower Apgar scores, particularly at 5 min, is closely linked to a higher likelihood of IVH (21–23). This association likely stems from oxygen deprivation related to birth or severe respiratory and circulatory instability, which can trigger oxidative damage and weaken the blood-brain barrier by harming vascular endothelial cells. Additionally, such complications may disrupt the brain's ability to regulate blood flow, increasing the chances of hemorrhage in the germinal matrix (21). These insights highlight the importance of effective resuscitation techniques and close postnatal observation to minimize IVH risk in newborns.

Septic shock and pulmonary bleeding represent intricate medical scenarios that can synergistically contribute to IVH via multiple biological pathways. Research by Luca and colleagues demonstrated a clear link between sepsis—whether confirmed by culture or clinically suspected—and IVH in premature newborns. Advanced statistical modeling revealed sepsis as a standalone predictor of IVH (OR 4.7, 95% CI 1.7–13.1), with the danger becoming particularly pronounced when vasopressor therapy became necessary for circulatory collapse (24). The systemic inflammation characteristic of septic shock compromises the brain's protective barrier, enabling toxic elements to penetrate neural tissue. Simultaneously, the lack of proper control over blood flow to the brain in premature babies makes the white matter particularly susceptible to harm, which can result in issues like cystic periventricular leukomalacia or widespread, non-cystic white matter damage (25, 26). Furthermore, interventions required for sepsis management—such as mechanical ventilation or catecholamine administration—can induce fluctuations in both intrathoracic and intracranial pressures, destabilizing cerebral perfusion (22). Systemic inflammatory response syndrome triggered by sepsis ramps up the production of inflammatory signaling molecules like IL-6 and TNF-α, while simultaneously kicking the coagulation system into overdrive. This cascade of events throws off the body's careful equilibrium between clot formation and breakdown, frequently leading to a dangerous depletion of platelets and the development of disseminated intravascular coagulation—a double whammy that severely compromises the blood's ability to clot properly (27). Pulmonary hemorrhage, frequently accompanied by hypoxemia and hemodynamic instability, indirectly exacerbates fluctuations in cerebral perfusion pressure and increases vascular shear stress, further compromising vascular integrity. These pathological processes are consistent with the mechanisms proposed in neonatal sepsis associated brain injury research.

Our study found that decreased hemoglobin and thrombocyte counts are associated with IVH, emphasizing the pivotal roles of anemia and coagulopathy in its pathogenesis. Christine et al. identified low initial hematocrit as a significant factor linked to IVH, incorporating it as an independent risk variable in their multivariable model (27). Similarly, Siddappa et al. reported that infants with severe IVH exhibited prolonged prothrombin time and international normalized ratio, along with lower thrombocyte counts (28). Furthermore, a predictive model developed by Weinstein et al. demonstrated that higher hematocrit and thrombocyte levels were significantly associated with a reduced incidence of moderate to severe IVH (2). These findings are consistent with previous studies and suggest that maintaining hematologic parameters within normal or elevated ranges may provide a protective effect, while lower levels are linked to an increased risk of IVH.

Given the serious implications of IVH, especially in its more severe stages, creating reliable methods to predict which newborns are at highest risk is crucial for clinical practice. Kumar and colleagues introduced a simple scoring system that factors in gestational age, birth weight, Apgar score at 1 min, place of birth (inborn vs. outborn), and sex—yielding an AUC of 0.77 (29). Meanwhile, Weinstein's team developed a more refined model that includes gestational age, 5-min Apgar score, hemoglobin levels, and platelet count, achieving a stronger predictive accuracy with an AUC of 0.826 (2). Kim et al. employed machine learning algorithms (e.g., XGBoost) and demonstrated that early postnatal data could enhance predictive accuracy (30). However, existing models vary in terms of variable selection, prediction windows, and interpretability. This study employed a comprehensive LASSO regression approach that incorporated all candidate variables for initial screening. This comprehensive strategy capitalizes on LASSO's inherent advantage in handling high dimensional data by automatically shrinking coefficients of irrelevant variables to zero, thereby minimizing subjective selection bias. The selected predictors were subsequently validated through multivariable logistic regression to ensure robustness. The resulting nomogram for predicting early-onset severe IVH in extremely preterm infants demonstrates both strong predictive accuracy and clinical utility. Our predictive model combines six routinely available clinical variables—gestational age, 5-min Apgar score, septic shock, pulmonary hemorrhage, hemoglobin level, and platelet count—to assess severe IVH risk during the critical postnatal period. The model showed significant clinical utility (DCA net benefit range: 0.05–0.8), supporting its use for risk stratification in NICUs. Clinically, this tool enables: (1) intensified neuromonitoring (including serial cranial ultrasounds) for infants with >30% predicted risk, and (2) personalized interventions (hemodynamic support and transfusion threshold adjustment) for those exceeding 50% risk. By integrating real-time laboratory values with clinical indicators, the model allows for dynamic risk assessment that can be readily incorporated into NICU workflows. While demonstrating immediate clinical applicability, further validation across diverse settings and evaluation of long-term neurodevelopmental outcomes remain important next steps.

This research has a few notable drawbacks. First, the single center retrospective design and limited sample size may introduce selection bias and restrict the generalizability of findings. Our model's events-per-variable ratio (EPV = 7.7) falls below the conventional threshold of 10, though we mitigated this through regularization techniques and internal validation. While the 7:3 data split was justified for our cohort, external validation in multicenter studies with larger samples is needed to verify transportability across diverse clinical settings. Additionally, some potentially significant predictors like patent ductus arteriosus were not incorporated, and the study did not assess long term neurodevelopmental outcomes such as cerebral palsy or cognitive deficits. Future research should integrate longitudinal follow-up data to evaluate the model's predictive value for both short term and long term outcomes.

In conclusion, this study introduces an intuitive and clinically applicable nomogram designed to accurately predict the likelihood of severe IVH in extremely premature newborns. Tailored for healthcare providers—particularly those in frontline pediatric care—this tool leverages readily available patient data to pinpoint infants at elevated risk, allowing for proactive surveillance and early medical intervention. By streamlining risk assessment and supporting personalized treatment plans, the model could significantly mitigate both fatal outcomes and lasting neurological complications, thereby enhancing long term prognosis for these fragile patients.