Development and verification of a risk prediction model for bronchopulmonary dysplasia in very low birth weight infants

Background To analyze the risk factors of bronchopulmonary dysplasia (BPD) of very low birth weight infants (VLBWIs), and to develop and verify a risk prediction model of BPD. Methods The data of 611 VLBWIs from the neonatal intensive care unit (NICU) of a tertiary grade A hospital in Suzhou from January 2017 to September 2019 were collected. The data was randomly divided into the modeling set (451 cases) and the validation set (160 cases). Binary logistic regression was used to analyze the data, and the model was examined by a receiver operating characteristic (ROC) curve. The grouped data was used to verify the sensitivity and specificity of the model. Results The study found that neonatal asphyxia, the positive rate of sputum culture, neonatal sepsis, neonatal respiratory distress syndrome (NRDS), blood transfusions (≥3), patent ductus arteriosus (PDA), the time of invasive mechanical ventilation, the duration of oxygen therapy, and the time of parenteral nutrition were the independent risk factors of BPD, while 1 min Apgar score was a protective factor. The model formula was Z=neonatal asphyxia * 1.229 + the positive rate of sputum culture * 1.265 + neonatal sepsis * 1.677 + NRDS * 1.848 + blood transfusions (≥3) * 1.455 + PDA * 1.835 - 1 min Apgar score * 0.25 + the time of invasive mechanical ventilation * 0.123 + the duration of oxygen therapy * 0.09 + the time of parenteral nutrition * 0.057 - 8.077. The area under the ROC curve of this model was 0.965 (95% CI: 0.946-0.983), with a sensitivity of 93.7% and a specificity of 91.3%. Verification of this prediction model showed a sensitivity of 92.9% and a specificity of 76%, demonstrating that the effects of this model were satisfactory. Conclusions The risk prediction model had a good predictive effect for the risk of BPD in VLBWIs, and can provide a reference for preventive treatment and nursing intervention.