To compare the cumulative incidence of infections acquired in the pediatric intensive care unit (PICU) and neonatal intensive care unit (NICU). Estimation of the cumulative incidence of infections with data obtained from the Pediatric Prevention Network (PPN) point-prevalence survey and observed rates from the National Nosocomial Infections Surveillance (NNIS) system. Ten hospitals participated in both the PPN survey and NNIS system. All patients present on the PPN survey dates (August 4, 1999, or February 1, 2000) in the NICUs or PICUs of the PPN hospitals were included in the survey. Point prevalences for PICU-acquired and for NICU-acquired infections at these hospitals were calculated from the survey data. The cumulative incidence rates were estimated from the point prevalence rates using a standard formula and a standard method for calculating the time to recovery (ie, on the basis of the assumption that discontinuance of antimicrobial therapy indicates recovery from infection); alternate methods to judge the time to recovery from infection were also explored. The average cumulative incidence of intensive care unit-acquired infection for NICUs and PICUs combined (all units), as measured by NNIS, was 14.1 cases per 100 patients; in comparison, the prevalence was 14.06 cases for 100 patients (median difference, -0.95 cases per 100 patients; 95% confidence interval, -4.6 to 5.0 cases per 100 patients), and the estimated cumulative incidence using the standard method of calculating the time to recovery was 13.8 cases per 100 patients (median difference, -1.5 cases per 100 patients; 95% confidence interval, -9.1 to 2.9 cases per 100 patients). Estimates of cumulative incidence using alternate methods for calculation of time to recovery did not perform as well (range, 4.9-100.9 cases per 100 patients). The average incidence density for all units, as measured by the NNIS system, was 6.8 cases per 1,000 patient-days, and the estimate of incidence density using the standard method of calculating the time to recovery was 3.6 cases per 1,000 patient-days (median difference, 4.3 cases per 1,000 patient-days; 95% confidence interval, 0.9 to 9.2 cases per 1,000 patient-days). Estimated incidence densities using alternate methods for determining recovery time correlated closely with observed incidence densities. In this patient population, the simple point prevalence provided the best estimate of cumulative incidence, followed by use of a standard formula and a standard method of calculating the time to recovery. Estimation of incidence density using alternate methods performed well. The standard formula and method may provide an even better estimate of cumulative incidence than does simple prevalence in general populations.