In 1997, the United States Centers for Disease Control and Prevention (CDC) published revised case definitions for legionellosis which eliminated the previously used category of "probable case" based on a single indirect fluorescence antibody (IFA) titer. This study evaluated the influence of revision on the case rates of legionellosis in Taiwan. From 1988 to 2002, 4615 patients with pneumonia were tested for legionellosis in our hospital. The testing methods included IFA assay for serum specimens and direct fluorescence antibody (DFA) assay for sputum specimens. Using the revised criteria, Legionnaires' disease (LD) was diagnosed by DFA in 27 cases and by IFA in 11 cases. The most common underlying conditions were cigarette smoking (44.7%), chronic obstructive pulmonary disease (28.9%) and corticosteroid use (26.3%). The clinical features were nonspecific, including fever (73.7%), dyspnea (63.2%), cough (63.2%) and leukocytosis (63.2%). The overall mortality rate was 18.4%, and the directly LD-attributable mortality rate was 10.5%. Nasogastric tube insertion, endotracheal intubation, congestive heart failure before the onset of LD, inappropriate antimicrobial therapy, respiratory failure and absence of fever during the LD course were significantly associated with LD-attributable mortality. Older age (>70 years) was not associated with higher mortality (p=0.053). Using the revised diagnostic criteria in our series, the positive rate of case identification by IFA was 0.26%, while use of the previous case definitions resulted in a positive rate of 7.6% (including probable and definitive cases). Recognition that the original CDC criteria of IFA titer >1:256 or elevation of IFA titer <4-fold in paired sera could not adequately define an LD etiology has led to a dramatic lowering of case rates among studies after the criteria revision in Taiwan and elsewhere. Assays that are faster, more sensitive and less technician dependent are needed to diagnosis this disease.