Conversion of an intensive care unit (ICU) from an open unit to isolation rooms permitted study of patient care practices, colonization and infection in both settings. Air sampling and observation of patient care practices included 99 of 410 open unit patients (168 patient-hours during nine months) and 68 of 1,022 isolation room patients matched on the basis of risk factors for infection and staff contact (113 patient-hours during 12 months). Number and type of interactions between staff and patients, and frequency of handwashing and its relationship to patient-staff interactions were recorded. All ICU patients were monitored daily for signs of and selected risk factors for infection, and material for culture for six surveillance organisms was obtained every four days. Numbers of persons interacting with a patient hour were 6.1 +/- 3.5 in the open units and 4.9 +/- 2.8 in the isolation rooms (0.05 less than P less than 0.10). Frequency of handwashing did not increase significantly in the unit providing convenient sinks, occurring in an observed to expected ratio of only 24 percent. Over-all rates of infection in the open unit and isolation rooms were 15.0 and 13.4, respectively. Half of the infections occurring in patients with complete cultures obtained on admission were caused by organisms colonizing the patient upon admission to the ICU. The isolation rooms did not appear to reduce nosocomial acquisition (P = 0.168, Mantel-Haenszel) of the six surveillance organisms. We conclude that many patient-staff interactions in an ICU are not followed by handwashing, and that the new unit design had no apparent effect upon the frequency of handwashing or over-all incidence of colonization and infection in the ICU.