(1091) Validation of the Neonatal Research Network Extremely Preterm Birth Outcome Tool.

The neonatal research network (NRN) estimator of outcomes is widely used to provide individualized counseling to patients at risk for extremely preterm birth. The calculator was based on data obtained from 2006 through 2012. Our objective was to evaluate whether the output of the calculator remains valid in a contemporary cohort.


Study Design:
This was a secondary analysis of a randomized trial of pessary in singleton pregnancies with short cervix recruited from 2017 to 2021. All live neonates delivered from 22 0/7 through 25 6/7 weeks’ with no congential malformations were included. Those with no resuscitation after birth were excluded. The predicted survival using the NRN estimator was calculated for each neonate. The mean predicted survival for the neonates within each gestational age week and also for selected characteristics was calculated with 95% confidence intervals around the mean. The predicted survival was compared to the actual percent survival using a One-Proportion test. A flexible calibration curve was produced using a non-parametric (loess) smoother to assess how well the predicted survival probabilities matched the actual survival.

Results:
Of 544 patients enrolled in the trial, 56 were included that delivered in the periviable period. The 42 neonates who survived were born later in gestation, had higher birthweight, and were more likely to have received corticosteroids compared with those who did not survive (Table 1). The actual average rate of survival per given week was slighty higher than what was predicted by the NRN model. The actual survival for those self-reported as NH Black was slightly higher than predicted survival (87% vs. 63%, p=0.01).The calibration curve showed that the calculator overestimated survival at lower actual survival rates, underestimated survival at higher survival rates, and had a strong prediction overall (ROC 0.75 95%CI 0.58, 0.87, Figure 1).

Conclusion:
Overall , the NRN estimator tool somewhat underestimates survival, but not to a significant degree and the tool estimates remain valid.