The CRIB (clinical risk index of babies) score was developed to overcome the disadvantages of birthweight-specific comparisons between neonatal units. The aims of this study were to assess the ability of CRIB score compared to birthweight and gestational age to predict hospital mortality in very low birthweight infants and to use CRIB score in auditing one unit's performance during a prolonged time period. The charts of 335 infants with birthweight < or = 1500 g born between 1980 and 1995 were reviewed retrospectively. CRIB predicted hospital mortality significantly better than birthweight and gestation and performed equally well, whether the infants were treated with synthetic surfactant or not. When adjusting for CRIB score there was a significant improvement in the unit's performance, probably owing to the introduction of surfactant. As small samples tend to be associated with wide confidence intervals, use of CRIB is recommended in comparing risk adjusted mortality in a single unit over several years, as in this study, or between large groups of neonatal units over shorter periods.