The performance of three scoring systems for assessing mortality risk for neonates--clinical risk index for babies (CRIB), score for neonatal acute physiology (SNAP), and SNAP's perinatal extension (SNAP-PE)--were tested in the same set of patients. In 222 neonates weighing less than 1500 g at birth, CRIB scores were significantly better for assessing mortality risk than SNAP (p = 0.017) or SNAP-PE (p < 0.001), areas under receiver operating characteristic curves being 0.89 (SE 0.02), 0.82 (0.03), and 0.79 (0.03), respectively. Male sex was independently associated with poor prognosis after taking the CRIB score into account with a risk ratio of 2.75. We conclude that CRIB is the most useful score for comparing the performance of neonatal intensive-care units. New treatment methods, however, may require modifications to the system.