[Establishment and efficiency test of a clinical prediction model of bronchopulmonary dysplasia associated pulmonary hypertension in very premature infants]
J K Cao,H Q Fan,Y B Xiao,D Wang,C G Liu,X M Peng,X R Gao,S H Tang,T Han,Y B Mei,H Y Liang,S M Wang,F Wang,Q P Li
DOI: https://doi.org/10.3760/cma.j.cn112140-20230912-00178
2024-02-02
Abstract:Objective: To develop a risk prediction model for identifying bronchopulmonary dysplasia (BPD) associated pulmonary hypertension (PH) in very premature infants. Methods: This was a retrospective cohort study. The clinical data of 626 very premature infants whose gestational age <32 weeks and who suffered from BPD were collected from October 1st, 2015 to December 31st, 2021 of the Seventh Medical Center of the People's Liberation Army General Hospital as a modeling set. The clinical data of 229 very premature infants with BPD of Hunan Children's Hospital from January 1 st, 2020 to December 31st, 2021 were collected as a validation set for external verification. The very premature infants with BPD were divided into PH group and non PH group based on the echocardiogram after 36 weeks' corrected age in the modeling set and validation set, respectively. Univariate analysis was used to compare the basic clinical characteristics between groups, and collinearity exclusion was carried out between variables. The risk factors of BPD associated PH were further screened out by multivariate Logistic regression, and the risk assessment model was established based on these variables. The receiver operating characteristic (ROC) area under curve (AUC) and Hosmer-Lemeshow goodness-of-fit test were used to evaluate the model's discrimination and calibration power, respectively. And the calibration curve was used to evaluate the accuracy of the model and draw the nomogram. The bootstrap repeated sampling method was used for internal verification. Finally, decision curve analysis (DCA) to evaluate the clinical practicability of the model was used. Results: A total of 626 very premature infants with BPD were included for modeling set, including 85 very premature infants in the PH group and 541 very premature infants in the non PH group. A total of 229 very premature infants with BPD were included for validation set, including 24 very premature infants in the PH group and 205 very premature infants in the non PH group. Univariate analysis of the modeling set found that 22 variables, such as artificial conception, fetal distress, gestational age, birth weight, small for gestational age, 1 minute Apgar score ≤7, antenatal corticosteroids, placental abruption, oligohydramnios, multiple pulmonary surfactant, neonatal respiratory distress syndrome (NRDS)>stage Ⅱ, early pulmonary hypertension, moderate-severe BPD, and hemodynamically significant patent ductus arteriosus (hsPDA) all had statistically significant influence between the PH group and the non PH group (all P<0.05). Antenatal corticosteroids, fetal distress, NRDS >stage Ⅱ, hsPDA, pneumonia and days of invasive mechanical ventilation were identified as predictive variables and finally included to establish the Logistic regression model. The AUC of this model was 0.86 (95%CI 0.82-0.90), the cut-off value was 0.17, the sensitivity was 0.77, and the specificity was 0.84. Hosmer-Lemeshow goodness-of-fit test showed that P>0.05. The AUC for external validation was 0.88, and the Hosmer-Lemeshow goodness-of-fit test suggested P>0.05. Conclusions: A high sensitivity and specificity risk prediction model of PBD associated PH in very premature infants was established. This predictive model is useful for early clinical identification of infants at high risk of BPD associated PH.