Establishment And Efficacy Test Of Clinical Prediction Model Of BPD-PH In Very Preterm Infants

Objective:Develop and validate a risk prediction model for identifying bronchopulmonary dysplasia associated pulmonary hypertension(BPD-PH).To provide effective clinical guidance for preventing the occurrence of BPD-PH and improving the prognosis of very premature infants.Methods:The clinical data of premature infants whose gestational age were<32 weeks and who suffered from bronchopulmonary dysplasia(BPD)were collected from October 1,2015 to December 31,2021 in the neonatal ward 1 and Ward 2,Department of Pediatrics,the Seventh Medical Center of PLA General Hospital.The clinical information includes maternal information during pregnancy and neonatal information.According to the results of the Doppler echocardiography over 28 days after birth,the infants were divided into two groups:BPD-PH group and non BPD-PH group.Univariate analysis was used to compare the basic clinical characteristics between groups,and collinearity exclusion is carried out between variables.The influencing factors of BPD-PH were screened out by multivariate logistic regression for statistically significant clinical variables,and the risk assessment model was built according to these variables.We used the receiver operating characteristic(ROC)area under curve(AUC),Hosmer-Lemeshow test and calibration curve to evaluate the model’s differentiation and fit.The bootstrap repeated sampling method was used for internal verification.Clinical information of very premature infants with BPD in the neonatal ward of Hunan Children’s Hospital from January 1,2020 to December 31,2021 was collected for external verification.Finally,we drew a nomogram of the model and used decision curve analysis(DCA)to evaluate the clinical practicability of the model.Results:1.A total of 626 patients were included for modeling and internal verification,including 541 patients in the non BPD-PH group and 85 patients in the BPD-PH group.A total of 229 patients were included for external verification,including 24 patients with BPD-PH and 205 patients without BPD-PH.2.Univariate analysis of the modeling set showed statistical differences in artificial conception,fetal distress,gestational age,birth weight,1 minute Apgar score ≤7,antenatal corticosteroids,a full course of antenatal corticosteroids,placental abruption,oligohydramnios,multiple PS use,neonatal respiratory distress syndrome(RDS)>stage Ⅱ,early pulmonary hypertension,moderate-severe BPD,and hemodynamically significant patent ductus arteriosus(hsPDA),PDA ligation,sepsis,pulmonary hemorrhage,intracranial hemorrhage≥Ⅲ,severe retinopathy of prematurity,pneumonia,days of invasive mechanical ventilation,and high frequency ventilation.(all P<0.05)Predictive variables were screened by multivariate Logistic regression analysis,and antenatal corticosteroids,a full course of antenatal corticosteroids,,fetal distress,RDS>stage Ⅱ,hsPDA,pneumonia and days of invasive mechanical ventilation were finally included to establish Logistic regression model.3.The AUC of this model was 0.86(95%CI:0.818-0.902),the cut-off value was 0.16,the sensitivity was 75.0%,and the specificity was 83.2%.Hosmer-Lemeshow test showed that P>0.05,indicating good model differentiation and calibration.Conclusion:A preliminary prediction model for the occurrence of BPD-PH in very premature infants was established.This risk assessment model can help identify infants at high risk of BPD-PH,and help guide personalized treatment for BPD-PH prevention and improve prognosis.4.Internal verification results indicate that the model is consistent well.The AUC for external validation was 0.90,and the Hosmer-Lemeshow test suggested P>0.05.Draw and display a nomogram of the model according to the regression equation.