A six-month prospective survey was carried out in a university hospital to assess the incidence of Acinetobacter baumannii cross-contamination and to identify risk factors for colonization. Clinical isolates obtained during the study period were biotyped and genotyped by pulsed-field gel electrophoresis after ApaI macrorestriction of total DNA. Case-control univariate and multivariate analyses were performed to identify risk factors for Acinetobacter baumannii colonization. One hundred forty-seven patients hospitalized in 36 units were colonized or infected, of whom 52 were in three intensive care units. The urinary (29%) and bronchopulmonary tracts (26%) were the most frequently colonized sites. Nine major restriction patterns were identified: two were exhibited by epidemic multi-resistant strains of biotype 9 which were isolated from 65 patients hospitalized in ten units. Multivariate analysis showed that case-patients were (a) more likely than non-infected controls to be male, to have been previously hospitalized in another unit and to have had longer stays in the unit before colonization and hyperalimentation; and (b) more likely than controls colonized with other gram-negative bacilli to be male, to have had longer hospitalization, to have received treatment with third-generation cephalosporins and to have had a urinary catheter. The high incidence of colonization with Acinetobacter baumannii can thus be attributed to frequent cross-contamination and the use of broad-spectrum antibiotics. Colonized patients appear to be the major source of cross-contamination as epidemic strains spread throughout the hospital.