Abstract Backround: Pulmonary hemorrhage (PH) is a serious lung disease that endangers the safety of premature infants, but its occurrence is difficult to predict.Our research aims to establish and validate a predictive model for pulmonary hemorrhage in extremely premature infants. Methods:This single-center retrospective cohort study included 402 preterm infants admitted to the NICU of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, between June 1, 2018, and October 31, 2023. Participants were randomly allocated into a training set (n=277) and a validation set (n=125).Univariate and multivariate logistic regression analyses identified independent risk factors. A nomogram prediction model was constructed and evaluated using ROCcurves, calibration plots, and decision curve analysis (DCA). Result:Among 402 infants, 59 developed PH (incidence: 14.7%). Independent risk factors included hemodynamically significant patent ductus arteriosus (hsPDA; OR=2.62, P=0.019), invasive respiratory support on the first postnatal day (OR=3.86, P=0.013), sepsis (OR=4.56, P=0.004), and shock (OR=4.25, P=0.005). The model demonstrated excellent predictive performance in both training and validation sets (AUC: 0.858 vs. 0.879). Hosmer-Lemeshow tests (training set P=0.957; validation set P=0.738) and calibration curves indicated good consistency. DCA confirmed significant clinical net benefits. Conclusion:The developed nomogram provides a validated, visually accessible tool for early identification of high-risk preterm infants, enabling timely interventions to improve outcomes.
Qi et al. (Wed,) studied this question.