36 patients with severe community-acquired pneumonia, treated in an intensive care unit (ICU), were examined in a prospective study using a comprehensive diagnostic program to establish an early etiological diagnosis. The resulting prompt and adequate antimicrobial therapy may have decreased the number of fatal cases. Special emphasis was placed on the use of a method incorporating fiberoptic bronchoscopy, together with protected brush sampling and bronchial lavage. An etiological diagnosis was established in 81% (29/36) of the cases. This etiological diagnosis was established within 48-72 h in 53% (19/36) of the patients, S. pneumoniae being the most frequent agent found (12 patients). This information, however, was poorly utilized since in only 11/19 of these patients was the antimicrobial therapy changed from a broad-spectrum antibiotic to a more specific narrow spectrum agent. The overall mortality rate was 22% (8/36). 7/8 patients who died had compromising factors. Most deaths in community-acquired pneumonia are still associated with pneumococcal infection. We conclude that fiberoptic bronchoscopy with brush samples via a plugged double lumen catheter provides the least misleading information concerning the etiological agent in pneumonia; sampling should be done as soon as possible after admission to the hospital, ideally before the need for ICU treatment; factors other than prompt antimicrobial therapy may influence the outcome of severe community-acquired pneumonia.