This cross-sectional and prospective one year study evaluated adults admitted to an inner city hospital with community-acquired pneumonia. The study used extensive diagnostic methods to evaluate the etiologies of community-acquired pneumonia in hospitalized patients with differing immunologic status. Of 385 study patients, concurrent problems associated with immunosuppression were noted in 221 (57%) patients, 180 of whom were human immunodeficiency virus (HIV)-infected. The five most common causes of community-acquired pneumonia were: Streptococcus pneumoniae, Pneumocystis carinii, aspiration, Hemophilus influenzae, and gram-negative bacilli. Only 8.3% of patients had either Legionella, Chlamydia pneumoniae or Mycoplasma pneumoniae. Despite use of state-of-the-art diagnostic techniques, no diagnosis was made in 46 of 180 (25.6%) HIV-infected patients, 56 of 164 (34.1%) immunocompetent patients, and 20 of 41 (48.8%) non-HIV-infected immunosuppressed patients. The diagnostic yield of pre-antibiotic sputum culture for conventional bacteria was 99/155 (63.9%) compared to 52 of 169 patients (32.7%) with adequate post-antibiotic sputum culture (p < 0.0001). Although S. pneumonia continues to be the most commonly identified etiologic agent of community-acquired pneumonia, it is surpassed by P. carinii in the HIV-infected patient population. The apparent decline in the frequency of S. pneumoniae in our series presumably reflects administration of antibiotics prior to procurement of sputum culture. The paucity of atypical agents in this study support the current American Thoracic Society guidelines for selective use of macrolide therapy in immunocompetent adults hospitalized with community-acquired pneumonia.(ABSTRACT TRUNCATED AT 250 WORDS)