We calculated the incidence of nosocomial infection in 2 intensive care units (ICUs) on the basis of prevalence data recorded from 1997 through 2002 and compared these estimates to cumulative incidences measured in the 2 ICUs during the same period to investigate the feasibility and the reliability of converting prevalence data to incidence estimates. Decreases in the calculated and measured incidences over time in the ICUs were found to be statistically significantly related.