Obstetric and perinatal data for 387 very low birthweight infants (less than or equal to 1500 g) who were admitted to a neonatal intensive care unit were used to derive and test a multivariate statistical model for predicting shortly after birth the risk of death before hospital discharge. Using the gestational age, the Apgar score at 5 minutes and the presence or absence of respiratory distress, the model correctly predicted outcome for 94% of survivors but for only 53% of deaths. The model successfully identified a subgroup with a low (less than 1 in 20) predicted risk of death. This subgroup comprised over one-third of the infants but included 47% of the survivors and only 5% of the deaths. The probabilities derived from the model are tabulated and may help paediatricians in maternity units without neonatal intensive care facilities to decide which very low birthweight infants to transfer for additional care.