To determine whether the Healthy People 2000 objective to deliver very-low-birthweight (VLBW) infants at subspecialty perinatal care centres was met, and if improvements in the regional perinatal care system could reduce neonatal mortality further for 2010, we examined place of delivery for VLBW infants, associated maternal characteristics and the potential impact on neonatal mortality. We used linked birth and death records for the 1994–96 Georgia VLBW (i.e. 500–1499 g) birth cohorts. Among 4770 VLBW infants, 77% were delivered at hospitals providing subspecialty perinatal care. The strongest predictor of birth hospital level was the mother's county of residence, defined using three levels: residence in a county with a subspecialty hospital, residence in a county adjacent to one with such a hospital or residence in a non-adjacent county. Eighty-nine per cent of infants born to women who resided in counties with subspecialty care hospitals delivered at such hospitals, compared with 53% of infants born to women who resided in a non-adjacent county. Women were also more likely to deliver outside subspecialty care if they had less than adequate prenatal care [adjusted odds ratio (AOR) 1.5, P-value = 0.0001]. The neonatal mortality rate varied by level of perinatal care at the birth hospital from 132.1/1000 to 283/1000 live births, with the highest death rate for infants born at hospitals offering the lowest level of care. Assuming that the differences in mortality were due to care level of the birth hospital, potentially 16–23% of neonatal deaths among VLBW infants could have been prevented if 90% of infants born outside subspecialty care were delivered at the recommended level. These findings suggest that a state's support of strong, collaborative, regional perinatal care networks is required to ensure that high-risk women and infants receive optimal health care. Improved access to recommended care levels should further reduce neonatal mortality until interventions are identified to prevent VLBW births.