The study was conducted at a tertiary care and teaching hospital with about 200 beds for internal medicine. The objective was to determine the diagnostic yield and value in directing antibiotic therapy of a routine microbial approach in patients with community-acquired pneumonia referred to a tertiary care center. We studied 93 episodes in a retrospective study. 69/93 (74%) cases were treated with at least one empirical antibiotic therapy prior to admission. Microbial investigation was performed in 83/93 cases (89%). An etiological agent was established in 19/83 (23%) cases including 7/50 (14%) by blood culture and 12/52 (23%) by serology. Bronchoscopy with 18 protected specimen brush and 20 bronchoalveolar lavage examinations was definitely diagnostic in only 1/25 (4%) cases, and this case was also identified by blood culture. 5/25 (20%) were probably diagnostic. Three pathogens, Streptococcus pneumoniae, Mycoplasma pneumoniae and Legionella pneumophila, accounted for 15/19 (79%) of the identified agents. The diagnostic results directed a change in antibiotic therapy in 6/19 (32%) of cases with definitely proven pathogens. 4/19 (21%) of cases would have been treated with an inappropriate regimen without diagnostic results. The diagnostic yield of routine microbial investigation in pretreated patients is low. The routine approach reveals its limited value especially in patients with severe courses. The role of bronchoscopy remains to be defined for patients with severe (and pretreated) community-acquired pneumonia.