We wished to develop a cost-effective, rapid strategy to detect and identify Bartonella species in the clinical laboratory and to determine the prevalence of Bartonella infection in the Houston veteran population. Bartonella colonies were identified by colony morphology, Gram stain, RapID ANA, repetitive extragenic palindromic-PCR (REP-PCR) and whole-cell fatty acid (CFA) analysis, and these methods were compared for their usefulness. A new test order for "Rochalimaea culture" (the genus Bartonella was previously known as the genus Rochalimaea) was instituted, and in addition, all blood specimens submitted for fungal culture (obtained in an isolator tube) were processed for Bartonella culture. Over a 16-month period we isolated Bartonella henselae from only 0.4% (2 of 533) of total cultures but from 1% (2 of 204) of human immunodeficiency virus-positive patients. After sufficient growth, identification of the Bartonella isolates to the species level could be obtained in 2 days. The REP-PCR allowed discrimination of all known species, whereas CFA analysis distinguished all except B. henselae and Bartonella quintana. The RapID ANA results failed to differentiate between B. henselae and B. quintana, and results for other species differed by only one or two tests. Blood obtained from a kitten which had been introduced into the household of one patient 2 months before the onset of fever yielded a Bartonella strain which was shown to be different from the strain from the patient and distinct from other Bartonella species by a combination of REP-PCR, CFA, and growth characteristics. Subsequent analysis of the citrate synthase gene sequence showed only an 86% similarity with any of the other known Bartonella species, suggesting that this isolate represents a distinct, previously uncharacterized species of Bartonella.