The aim of this prospective study was to assess the relative epidemiological role of digestive tract colonization by Acinetobacter baumannii, in comparison with other body site colonizations, in patients admitted to intensive care units (ICUs). From January to May 1995, axillary, pharyngeal and rectal swabs were taken together within the first 48 h of admission, and then weekly during ICU stay. Seventy-three patients were included, 48 of them (66%) had axillary, pharyngeal, or rectal colonization with A. baumannii, nine (19%) of these 48 during the first 48 h and the remaining 28 (77%) during the first week. Twenty-one (29%) had clinical samples positive for A. baumannii and axillary, pharyngeal, or rectal colonization. In 15 of these 21 (71%), colonization on body sites occurred prior to isolation from clinical samples (mean seven days, range 1-20). Throughout admission, rates of detection of A. baumannii were 75% (36/48) for axillary or pharyngeal swabs and 77% (37/48) for rectal swabs. Combination of two body site swabs yielded culture positive rates of 90% (43/48) for axillary-pharyngeal or axillary-rectal sites, and 96% (46/48) for pharyngeal-rectal. Two epidemic clones were defined by antibiotype and pulsed-field gel electrophoresis (PFGE) of SmaI DNA digests in 48 isolates from 11 patients. We conclude that body sites of patients were a major reservoir for A. baumannii infections in the outbreak. This finding cases doubt on the value of selective decontamination of the digestive tract as an additional infection control measure in this kind of outbreak. The weekly performance of pharyngeal and rectal swabs appears to detect A. baumannii colonization early among ICU patients and enables barrier methods to be applied rapidly.