Risk-adjustment may overcome the disadvantages of birthweight-specific comparisons of neonatal units. Risk-adjusted rates of death or impairment after 18 months were compared in five tertiary and three non-tertiary neonatal units for 695 high-risk infants. CRIB (clinical risk index for babies) was more closely related than birthweight to death or impairment. After adjustment for risk with CRIB, deaths remained more likely after non-tertiary than tertiary care (odds ratio 1.90, 95% CI 1.1-3.3) but rates of impairment in survivors were nearly identical (0.97, 0.5-2.1). To improve performance further, risk-adjusted rates of death and impairment should be monitored for all neonatal units.