After the death of a premature infant from rotavirus-associated necrotizing enterocolitis, we instituted prospective surveillance for this disease in our neonatal intensive care unit. During the 4-month study period an additional six cases of necrotizing enterocolitis and eight cases of hemorrhagic gastroenteritis occurred. Rotavirus infection was documented in 11 of these 15 symptomatic infants, in comparison with only eight rotavirus infections in 147 asymptomatic or minimally symptomatic bables (P<0.0001). Stools from 110 nursery personnel tested during the outbreak did not contain rotavirus. However, 12 of 59 staff members had serum IgM antibody against rotavirus, suggesting recent infection. In a case-control study we compared babies with severe gastrointestinal illness with a control group randomly selected from asymptomatic babies in the nursery during the time of the outbreak. Univariate analysis found six categorical variables and nine continuous variables that were significantly associated with disease. Multivariate logistic regression analysis, however, found only birth weight (P<0.0001), rotavirus infection (P<0.0001), and age at time of first nonwater feeding (P<0.02) to be associated with gastrointestinal illness. This study provides further evidence for the role of infection in some cases of neonatal necrotizing enterocolitis and hemorrhagic gastroenteritis.