We analyzed population-based data from the Georgia Neonatal Surveillance Network from 1974 to 1978 to determine the effect of the hospital of delivery on the neonatal mortality rate (NMR) of infants later admitted to neonatal intensive case units (NICUs). The NMR of 3,524 infants transported from primary centers to NICUs was significantly higher [relative risk (RR) = 2.1; 95% CL = 1.9 to 2.3] than that of 10,764 infants born in tertiary centers and admitted directly to an NICU. This effect persisted even after adjustment for birth-weight differences (RR = 1.6; 95% CL = 1.5 to 1.8). The relative risk in favor of delivery in a tertiary center increased with increasing birth weight. However, the proportion of infants of less than 1,000 gm surviving was higher for transported infants and increased with distance transported. This finding suggests that, at these very low birth weights, the hardiest infants were selectively transported. A surprising finding was the very low proportion of low-birth-weight infants delivered in primary centers and transported to NICUs (eg, 32% of infants weighing between 1,001 and 1,200 gm). Our findings support previous reports that delivery in a hospital with an NICU is preferable to later neonatal transport and suggest that efforts to increase the rate of maternal transport in high-risk pregnancies can lead to a substantial reduction in infant deaths.