In this issue of Pediatrics, Horbar and his colleagues¹ report on variations in mortality across 62 of the neonatal intensive care units (NICUs) participating in the Vermont Oxford Network in 1991 and/or 1992. The fact that NICUs experience substantial variation in their mortality and complication rates in ways that are not explained by the distribution of the basic characteristics of the patient population like birth weight, gestational age, race, and gender is now a well-documented phenomenon.^2-4 The difficulty has been in establishing the source of this variation. In this report, the investigators examine two potential sources of variation: the volume of patients as a proxy for experience, and presence of a training program as a marker of a teaching hospital. The concern about these particular factors reflects the rapid increase in the number of smaller NICUs in community hospitals outside the traditional large tertiary centers.⁵ Although they find significant variability in mortality controlling for a broader array of risk factors, neither volume nor the presence of a residency program influences this variation.

An older literature supports the conclusion of a survival advantage to infants born in a hospital with an NICU compared with hospitals without such units,^6-8 but comparisons among hospitals with NICUs is much more recent. In this very limited literature, the results obtained by Horbar et al¹ are somewhat at variance with two prior studies. In the study of Tarnow-Mordi and his colleagues,⁹ teaching hospitals perform better than nonteaching hospitals. The latter also appear to be relatively low-volume hospitals . . . [Full Text of this Article]