To control infections with endemic methicillin-resistant Staphylococcus aureus (MRSA) in a neonatal intensive care unit (NICU), triple dye was applied to the umbilical cords of infants in the intermediate-care but not the intensive-care area. The rate of MRSA infection, adjusted for time and intensity of care, decreased in the intermediate-care area (rate ratio, 0.35; 95% confidence interval [CI], 0.14-0.87; P < .01) but not in the intensive-care area (rate ratio, 0.92; 95% CI, 0.41-2.24; P = .48). After 22 months, the rate increased in both areas (Mantel-Haenszel rate ratio, 1.7; 95% CI, 1.0-2.8; P < .05) after overcrowding and understaffing increased. After temporary reduction of overcrowding and understaffing, extension of triple dye use to the intensive-care area and dedication of an infection control nurse to the NICU, MRSA colonization and infection rates decreased to near zero in both areas (infection rate ratios, 0.09 and 0.11, respectively; P < .005). The endemic MRSA strain, identified by pulsed-field gel electrophoresis, was eradicated.