To limit nosocomial spread of respiratory syncytial virus (RSV) infection, a longitudinal intervention trial was instituted. Nasal secretions or washes were screened for RSV antigen by enzyme-linked immunosorbent assay, and patients were assigned to an RSV-infected or an RSV-uninfected cohort. The baseline (preintervention) rate of 7.17 nosocomial cases of RSV per 1000 patient-days of care was used for comparison. Despite continued infections in the community after screening was initiated, there were no cases of RSV infection in 1880 patient-days of care for 3 months (p = 0.039). During the fourth month, an RSV-infected child was erroneously assigned to the RSV-uninfected cohort, and three nosocomial cases occurred--5.33/1000 patient-days of care (p = 0.286). Overall, there were three nosocomial RSV infections in 2443 patient-days of care in the 1987 season after screening was introduced--1.23/1000 patient-days of care (p = 0.026). In the subsequent RSV season, there was one nosocomial case--0.461/1000 patient-days of care for 3 months (p = 0.0074). During the same period, nosocomial cases of RSV were observed in the pediatric and neonatal intensive care units, where assignment to a cohort was not possible. We conclude that entry into a cohort at the time of admission, on the basis of prospective RSV screening by enzyme-linked immunosorbent assay, effectively reduces nosocomial transmission of RSV.