Advertisement

175: Predicting survival among neonates born at 23 weeks gestation

      Objective

      To evaluate the model derived from the Neonatal Research Network Extremely Preterm Birth Outcome Data in its ability to predict survival in a cohort of neonates born at 23 weeks gestation.

      Study Design

      We identified all neonates born at 23 0/7 to 23 6/7 weeks gestation who were resuscitated at our institution from January 1995 through December 2006. Data regarding gestational age at delivery, birth weight, neonatal gender, singleton versus multiple gestation, maternal administration of corticosteroids and survival to discharge were obtained from the medical record. The model recently published by Tyson et al., (N Engl J Med. 358:1672-81, 2008), was used to generate the predicted survival for individual neonates. The mean predicted survival for all infants defines the expected survival rate for the cohort. The observed rate of survival was compared to the expected survival rate for the cohort using the Chi-square test. The mean predicted survival for survivors versus nonsurvivors was also compared.

      Results

      We identified 33 neonates born at 23 0/7 to 23 6/7 weeks gestation who were resuscitated. Mean (± SD) birth weight was 559 gm (± 68.4). Maternal corticosteroids were administered in 15 (45.5%) of cases. Of the 33 infants, 16 (48.5%) were female, and 18 (54.5%) were singletons. The predicted survival rate for the cohort was 30.5%. The observed survival rate was 18/33 (54.5%), which was significantly higher than the predicted survival rate (p = 0.003). The mean predicted survival rate for survivors was 31.4%, versus 29.3% for nonsurvivors (p = 0.53).

      Conclusion

      The observed rate of survival to discharge among neonates delivered at 23 weeks gestation at our institution was significantly higher than the rate predicted using the published model, which may reflect institutional differences in selecting infants for resuscitation. There was no difference in predicted survival for survivors and nonsurvivors, suggesting that the published model may not be generalized to all populations.