Widespread use of fluconazole for the prophylaxis and treatment of candidiasis has led to a reduction in the number of cases of candidemia caused by Candida albicans but has also resulted in the emergence of candidemias caused by innately fluconazole-resistant, non-C. albicans Candida species. Given the fulminant and rapidly fatal outcome of acute disseminated candidiasis, rapid identification of newly emerging Candida species in blood culture is critical for the implementation of appropriately targeted antifungal drug therapy. Therefore, we used a PCR-based assay to rapidly identify Candida species from positive blood culture bottles. This assay used fungus-specific, universal primers for DNA amplification and species-specific probes to identify C. albicans, C. krusei, C. parapsilosis, C. tropicalis, or C. glabrata amplicons. It also used a simpler and more rapid (1.5-h) sample preparation technique than those described previously and used detergent, heat, and mechanical breakage to recover Candida species DNA from blood cultures. A simple and rapid (3.5-h) enzyme immunosorbent assay (EIA)-based format was then used for amplicon detection. One hundred fifty blood culture bottles, including 73 positive blood culture bottle sets (aerobic and anaerobic) from 31 patients with candidemia, were tested. The combined PCR and EIA methods (PCR-EIA) correctly identified all Candida species in 73 blood culture bottle sets, including bottles containing bacteria coisolated with yeasts and 3 cultures of samples from patients with mixed candidemias originally identified as single-species infections by routine phenotypic identification methods. Species identification time was reduced from a mean of 3.5 days by routine phenotypic methods to 7 h by the PCR-EIA method. No false-positive results were obtained for patients with bacteremias (n = 18), artificially produced non-Candida fungemias (n = 3), or bottles with no growth (n = 20). Analytical sensitivity was 1 cell per 2-microl sample. This method is simpler and more rapid than previously described molecular identification methods, can identify all five of the most medically important Candida species, and has the potential to be automated for use in the clinical microbiology laboratory.