Because of conflicting data about hospital-based transmission of Burkholderia (Pseudomonas) cepacia, an important respiratory pathogen in cystic fibrosis (CF), we compared strains found in sputum, lung, or blood of 29 CF patients in our center from 1988 to 1994, studying the relationship between strain and hospital exposure of incident and that of prevalent cases. Exposure was defined as a concurrent hospital stay between a prevalent and an incident case. B. cepacia strains were determined by polymerase chain reaction (PCR) ribotyping and endonuclease subtyping. The 16S to 23S spacer regions of the bacterial ribosomal RNA (rRNA) genes were amplified by PCR, and the product-size patterns used to type each B. cepacia isolate. Endonuclease digestion of the PCR products provided length polymorphisms for subtyping. There were 17 incident events during the period from 1988 to 1994, 16 of which involved a single ribotype. These 16 ribotypes could be divided into five subtypes by endonuclease mapping. Four patients grew B. cepacia from the blood, with the organism being the same strain as found in the lung in each case. Case controls were obtained to evaluate risk factors for B. cepacia acquisition. Concurrent hospitalization with a prevalent case significantly increased the risk of acquisition. There was no association between length of hospitalization, length of exposure, or FEV1 and the risk of B. cepacia acquisition.