Background: Some neonatologists state that at the delivery of extremely premature infants they rely on "how the baby looks" when deciding whether to initiate resuscitation. Previous studies have reported poor correlation between early clinical signs and prognosis.
Objective: To determine if neonatologists can accurately predict survival to discharge of extremely premature infants on the basis of observations in the first minutes after birth.
Methods: We showed videos of the resuscitation of 10 extremely premature infants (<26 weeks' gestation) to attending neonatologists and fellows from the 3 major perinatal centers in Melbourne, Australia. Antenatal information was available to the observers. A monitor visible in each video displayed the heart rate and oxygen saturation of the infant. Observers were asked to estimate the likelihood of survival to discharge for each infant at 3 time points: 20 seconds, 2 minutes, and 5 minutes after birth. The predictive ability of observers was expressed as the area (95% confidence interval [CI]) under the receiver-operating-characteristic curve.
Results: Seventeen attending neonatologists and 17 neonatal fellows completed the study. Receiver-operating-characteristic curves were generated for the combined and individual groups. Observers' ability to predict survival was poor (combined results): 0.61 (95% CI: 0.54-0.67) at 20 seconds, 0.59 (95% CI: 0.52-0.64) at 2 minutes, and 0.61 (95% CI: 0.55-0.67) at 5 minutes. Level of experience did not affect the observers' accuracy of predicting survival.
Conclusion: Neonatologists' reliance on initial appearance and early response to resuscitation in predicting survival for extremely premature infants is misplaced.