This prospective case-control study sought to identify differences in presentation between Legionella pneumonia (LP) diagnosed by urinary antigen and bacteremic pneumococcal pneumonia (PP), with the aim of assessing the ability of physicians to recognize such differences at admission and validating the Community-Based Pneumonia Incidence Study (CBPIS) Group scoring system for LP diagnosis. Significant differences in presentation were found: male sex, previous receipt of beta-lactam therapy, and temperature >39 degrees C were positively associated with LP; purulent sputum, pleuritic chest pain, and previous upper respiratory tract infection were negatively associated with LP. Physicians considered Legionella to be the most likely diagnosis in 52 (64%) of 81 LP cases and in 8 (6%) of 136 PP cases. Initial administration of a macrolide and rifampin and requests for urinary antigen testing for Legionella at admission were significantly more frequent among patients with LP. Overall, the CBPIS score did not differentiate reliably between LP and PP. Although certain presenting clinical features may allow recognition of LP, it is difficult to express them in a reliable scoring system.