Establishment and validation of a nomograph model for prediction of bronchopulmonary dysplasia in very low birth weight infants born earlier than 32 weeks / 中华围产医学杂志

ABSTRACT

Objective:To investigate the risk factors of bronchopulmonary dysplasia (BPD) in very low birth weight (VLBW) infants with gestational age ≤32 weeks within 28 days after birth and to establish and validate the nomogram model for BPD prediction.Methods:We retrospectively chose VLBW infants with gestational age ≤32 weeks who survived to postmenstrual age (PMA) 36 weeks and were admitted to the neonatal intensive care unit of Peking University Third Hospital from January 2016 to April 2020 as the training cohort. BPD was diagnosed in accordance with the 2018 criteria. The clinical data of these infants were collected, and the risk factors of BPD were analyzed by Chi-square test, Mann-Whitney U test, and multivariate logistic regression, and a nomogram model was established. The area under the curve (AUC) of the receiver operating characteristic (ROC) curve was used to assess the predictive performance. Decision curve analysis (DCA) was constructed for differentiation evaluation, and the calibration chart and Hosmer-Lemeshow goodness of fit test were used for the calibration evaluation. Bootstrap was used for internal validation. VLBW infants with gestational age ≤32 weeks survived to PMA 36 weeks and admitted to Hebei Chengde Maternal and Child Health Hospital from October 2017 to February 2022 were included as the validation cohort. ROC curve and calibration plot were conducted in the validation cohort for external validation. Results:Of the 467 premature infants included in the training cohort, 104 were in the BPD group; of the 101 patients in the external validation cohort, 16 were in the BPD group. Multivariate logistic regression analysis showed that low birth weight ( OR=0.03, 95% CI 0.01-0.13), nosocomial pneumonia ( OR=2.40, 95% CI 1.41-4.09), late-onset sepsis ( OR=2.18, 95% CI 1.18-4.02), and prolonged duration of endotracheal intubation ( OR=1.61, 95% CI 1.26-2.04) were risk factors for BPD in these groups of infants (all P<0.05). According to the multivariate logistic regression analysis results, a nomogram model for predicting BPD risk was established. The AUC of the training cohort was 0.827 (95% CI 0.783-0.872), and the ideal cut-off value for predicted probability was 0.206, with a sensitivity of 0.788 (95% CI 0.697-0.862) and specificity of 0.744 (95% CI 0.696-0.788). The AUC of the validation cohort was 0.951 (95% CI0.904-0.999). Taking the prediction probability of 0.206 as the high-risk threshold, the sensitivity and specificity corresponding to this value were 0.812 (95% CI 0.537-0.950) and 0.882 (95% CI 0.790-0.939). The Hosmer-Lemeshow goodness-of-fit test in the training and validation cohort showed a good fit ( P>0.05). DCA results showed a high net benefit of clinical intervention in very preterm infants when the threshold probability was 5%~80% for the training cohort. Conclusion:Low birth weight, nosocomial pneumonia, late-onset sepsis, and prolonged tracheal intubation duration are risk factors for BPD. The established nomogram model has a certain value in predicting the risk of BPD in VLBW less than 32 weeks.