Attending Physician National Cheng Kung University Hospital Tainan, Tainan, Taiwan (Republic of China)
Background: The number of peri-viable infants (22–25 weeks’ gestation) has increased with advances in perinatal care. Despite this, these extremely preterm infants remain at high risk for mortality and severe morbidities, highlighting the need for accurate survival prediction and individualized prognosis. Objective: This study aimed to identify prenatal and perinatal risk factors for early neonatal mortality (on the 3rd, 7th, and 28th day of life) in peri-viable preterm infants, develop and validate a regression-based predictive model, and establish a web-based scoring system for mortality risk assessment. Design/Methods: This study used data from the Preterm Infants Foundation (2016–2019), including infants born before 37 weeks and weighing under 1500 g. The dataset was randomly divided into a training set (60%) and a validation set (40%). Baseline characteristics were compared, and univariate and multivariate logistic regressions were performed to identify factors associated with 28-day mortality. Receiver operating characteristic (ROC) curves evaluated model discrimination, which was further tested in the validation set. Cox regression was also conducted to analyze predictors of 28-day mortality. Results: There were no significant differences in baseline characteristics between the training and validation sets (Table 1). Univariate logistic regression identified antenatal corticosteroid use, antenatal magnesium sulfate exposure, gestational age, birth weight, 1- and 5-minute Apgar scores, base excess, pH, body temperature, and postnatal glucose as factors associated with 28-day mortality. Multivariate analysis showed that gestational age (especially 22–23 weeks), birth weight (per 100 g), 5-minute Apgar score, pH, and body temperature were independent predictors of 28-day mortality. The ROC analysis yielded an AUC of 0.8048 (sensitivity 75.9%, specificity 71.0%, accuracy 72.4%) in the training set and 0.7743 (sensitivity 71.8%, specificity 68.1%, accuracy 69.1%) in the validation set. The time-dependent AUCs were 0.8096 and 0.8047 for the training and validation sets, respectively. Cox regression confirmed gestational age (22 weeks), birth weight, 5-minute Apgar score, pH, and body temperature as significant predictors of 28-day mortality.
Conclusion(s): The predictive model demonstrated moderate accuracy, with gestational age, birth weight, 5-minute Apgar score, pH, and body temperature identified as key predictors of neonatal mortality. These findings highlight the importance of targeted clinical care and optimized resource allocation for peri-viable preterm infants.
