Isolating carriers of multidrug-resistant (MDR) Acinetobacter baumannii is the main measure to prevent its spread. Identification of carriers accompanied by contact precautions is essential. We aimed to determine the appropriate surveillance sampling sites and the duration of carriage of MDR A. baumannii. We studied prospectively two groups of patients from whom MDR A. baumannii was previously isolated: (i) those with recent clinical isolation (<or=10 days) and (ii) those with remote clinical isolation (>or=6 months). Screening for carriage was conducted from six sites: nostrils, pharynx, skin, rectum, wounds, and endotracheal aspirates. Strains recovered concurrently from different sites were genotyped using pulsed-field gel electrophoresis. Twelve of 22 with recent clinical isolation of MDR A. baumannii had >or=1 positive screening culture, resulting in a sensitivity of 55% when six body sites were sampled. Sensitivities of single sites ranged from 13.5% to 29%. Among 30 patients with remote clinical isolation, screening cultures were positive in 5 (17%), with a mean duration of 17.5 months from the last clinical culture. Remote carriers had positive screening cultures from the skin and pharynx but not from nose, rectum, wounds, or endotracheal aspirates. Eleven strains from five patients were genotyped. In all but one case, isolates from different sites in a given patient were clonal. Current methodology is suboptimal to detect MDR A. baumannii carriage. The sensitivity of surveillance cultures is low, even when six different body sites are sampled. The proportion of individuals with previous MDR A. baumannii isolation who remain carriers for prolonged periods is substantial. These data should be considered when designing measures to limit the spread of MDR A. baumannii.