Abstract
BACKGROUND: In China, the number of preterm infants is the second largest globally. Compared with those in developed countries, the mortality rate and proportion of treatment abandonment for extremely preterm infants (EPIs) are higher in China. It would be valuable to conduct a multicenter study and develop predictive models for the mortality risk. This study aimed to identify a predictive model among EPIs who received complete care in northern China in recent years. METHODS: This study included EPIs admitted to eighteen neonatal intensive care units (NICUs) within 72 hours of birth for receiving complete care in northern China between January 1, 2015, and December 31, 2018. Infants were randomly assigned into a training dataset and validation dataset with a ratio of 7:3. Univariate Cox regression analysis and multiple regression analysis were used to select the predictive factors and to construct the best-fitting model for predicting in-hospital mortality. A nomogram was plotted and the discrimination ability was tested by an area under the receiver operating characteristic curve (AUROC). The calibration ability was tested by a calibration curve along with the Hosmer-Lemeshow (HL) test. In addition, the clinical effectiveness was examined by decision curve analysis (DCA). RESULTS: A total of 568 EPIs were included and divided into the training dataset and validation dataset. Seven variables [birth weight (BW), being inborn, chest compression in the delivery room (DR), severe respiratory distress syndrome, pulmonary hemorrhage, invasive mechanical ventilation, and shock] were selected to establish a predictive nomogram. The AUROC values for the training and validation datasets were 0.863 [95% confidence interval (CI): 0.813-0.914] and 0.886 (95% CI: 0.827-0.945), respectively. The calibration plots and HL test indicated satisfactory accuracy. The DCA demonstrated that positive net benefits were shown when the threshold was >0.6. CONCLUSIONS: A nomogram based on seven risk factors is developed in this study and might help clinicians identify EPIs with risk of poor prognoses early.