In the past several years, prediction models for severe intraventricular hemorrhage (IVH) in premature infants have emerged. However, few models have considered the importance of predictors related to the clinical course and hemostatic profile in predicting the risk of hemorrhage, such as the FiO2, hematocrit, and platelet count. Moreover, it is noteworthy that most models unreasonably confuse late-onset IVH with early-onset, posing a high risk of bias. The present study was performed to construct a new prediction model for severe IVH. The data for this population-based study came from a children's hospital. After screening by inclusion and exclusion criteria, 1009 very low birth weight infants (VLBWIs) were subsequently recruited in the study and divided into training and validation sets in a ratio of 7:3. Gestational age, Max FiO2, hematokrit on admission < 45%, and platelet count on admission < 100 × 109/L were incorporated into the nomogram chart. The area under the curve (AUC) values demonstrated robust predictive performance, with the training set yielding an AUC of 0.884 (bootstrap-corrected AUC = 0.903) and the validation set achieving an AUC of 0.859. The Delong test showed no statistically significant difference in AUCs between the training set and validation set (p = 0.528). The result of the Hosmer-Lemeshow test showed the model is well calibrated (p = 0.757). The present study identified the predictor model associated with severe IVH during the first 7 days of life, and the nomogram performed soundly, which would be a promising tool for early stratification of the risk for severe IVH in VLBWIs.
Keywords: intraventricular hemorrhage; neonate; neurodevelopmental disorders; risk prediction model; very low birth weight.
© 2025 The Author(s). The Kaohsiung Journal of Medical Sciences published by John Wiley & Sons Australia, Ltd on behalf of Kaohsiung Medical University.