Objective To determine the accuracy of the National Institute of Child Health and Human Development (NICHD) calculator in predicting death and neurodevelopmental impairment in Australian and New Zealand infants. Design Population-based cohort study. setting Australia and New Zealand. Patients Preterm infants 22–25 completed weeks gestation. Interventions Comparison of NICHD calculator predicted rates of death and death or neurodevelopmental impairment, with actual rates recorded in the Australian and New Zealand Neonatal Network cohort. Main outcome measures Infant death and death or neurodevelopmental impairment rates. results A total of 714 infants were included in the study. Of these infants, 100 (14.0%) were <24 weeks, 389 (54.5%) male, 529 (74.1%) were singletons, 42 (5.9%) had intrauterine growth restriction, 563 (78.9%) received antenatal steroids and 625 (87.5 %) were born in a tertiary hospital. There were 288 deaths (40.3%), 75 infants (10.5%) with neurodevelopment impairment and 363 (50.8%) with death or neurodevelopmental impairment. The area under the curve (AUC) for prediction of death and the composite death or neurodevelopmental impairment by the NICHD calculator in our population was 0.65(95% CI 0.61 to 0.69) and 0.65 (95% CI 0.61 to 0.69), respectively. When stratified and compared with gestational age outcomes, the AUC did not change substantially for the outcomes investigated. The calculator was less accurate with outcome predictions at the extreme categories of predicted outcomes—underestimation of outcomes for those predicted to have the lowest risk (<20%) and overestimation for those in the highest risk category (≫80%). conclusion In our recent cohort of extremely preterm infants, the NICHD model does not accurately predict outcomes and is marginally better than gestational age based outcomes.