The decision to develop rotavirus vaccines was predicated on the extensive burden of rotavirus disease among children worldwide. US reports on nationwide hospitalizations (1979-1992) and deaths (1968-1991) due to diarrhea and weekly reports of rotavirus infection by 74 laboratories were reviewed to estimate the burden of rotavirus disease, identify epidemiologic trends, and consider methods for evaluating an immunization program when a vaccine becomes available. From 1968 to 1985, diarrhea-related deaths among US children <5 years old declined from 1100 to 300/year. This decline was associated with the disappearance of winter peaks for diarrhea-related deaths previously associated with rotavirus infection among children 4-23 months old. From 1979 to 1992, however, hospitalizations for diarrhea averaged 186,000/year and retained their winter peaks, which have been linked to rotavirus infections. Each year an estimated 54,000-55,000 US children are hospitalized for diarrhea, but <40 die with rotavirus. A rotavirus vaccine program will require improved surveillance, including the timely collection of data from sentinel hospitals, in which a diagnosis of rotavirus can be established or ruled out for all children hospitalized for diarrhea.