Early predictors of motor development impairments in preterm infants with bronchopulmonary dysplasia: A prospective cohort study

Background. Despite advancements in modern neonatal intensive care, the risk of developing permanent neurological disorders, including cerebral palsy, remains high among very preterm infants with bronchopulmonary dysplasia. Therefore, early risk prediction for cerebral palsy in this patient population is of critical importance. Objective. To identify prognostic factors for cerebral palsy in very preterm infants with bronchopulmonary dysplasia based on antenatal and neonatal data. Methods. A prospective cohort study of 71 preterm infants with bronchopulmonary dysplasia was conducted at the Neonatal Intensive Care and Follow-Up Departments of the Children’s Regional Clinical Hospital (Ministry of Health of the Krasnodar Krai). This was followed by a retrospective analysis of risk factors for cerebral palsy, with patient follow-up extending up to 24 months of corrected age. Based on the outcomes, the children were divided into two groups: Group I ( n = 23) comprised patients with cerebral palsy, and Group II ( n = 48) included those without it. The analysis was primarily focused on identifying factors that most significantly affect the risk of cerebral palsy, accounting for antenatal and neonatal characteristics, to subsequently develop a predictive model. To build this predictive model, logistic regression with L1 regularization was applied; internal validation was performed using nested repeated stratified cross-validation. Results. Between-group comparison revealed statistically significant differences in birth weight and gestational age. In Group I, the median birth weight was 1100.0 g (IQR 1050.0–1130.0) versus 1255.0 g (IQR 1140.5–1400.0) in Group II ( p < 0.001); the median gestational age was 29.0 (28.0–29.0) and 30.0 (29.0–31.0) weeks, respectively ( p = 0.002). No statistically significant differences were found regarding sex ( p = 0.485). As the most stable predictors, the predictive model identified intraventricular hemorrhage grade, birth weight, and chronic fetoplacental insufficiency. Following internal validation, the area under the ROC curve was 0.900 (95% CI: 0.817–0.961). At a classification threshold of 0.5, the model demonstrated a sensitivity of 73.9%, a specificity of 85.4%, and an accuracy of 81.7%. Conclusion. The developed predictive model enables early identification of very preterm infants with bronchopulmonary dysplasia who are at high risk for cerebral palsy, justifying early monitoring and rehabilitation approaches.