The aim of the study was to investigate the etiology and the impact of invasive quantitative sampling on the management of severe pneumonia in institutionalized older people with antimicrobial treatment failure. Fifty-two institutionalized patients aged 70 years and older hospitalized with a presumptive diagnosis of severe pneumonia and failure to respond to treatment after 72 hours of initiation of outpatient antimicrobial therapy were enrolled. Microbial investigation included blood culture, serology, pleural fluid, and bronchoalveolar samples. A definite etiology could be established in 24 of 52 (46%) patients. Methicillin-resistant Staphylococcus aureus (33%), enteric Gram-negative bacilli (24%), and Pseudomonas aeruginosa (14%) accounted for most isolates. Atypical infections (2%) were uncommon. Invasive bronchial sampling directed a change of microbial therapy in 8 (40%) and discontinuation of antibiotics in 2 of 20 cases (10%) of definite pneumonia. Overall hospital mortality was 42%. There was no difference in mortality among definite or unverified cases or those who had invasive bronchial sampling-guided change in therapy. We conclude that antimicrobial therapy should be targeted toward "nosocomial" pathogens in those institutionalized patients who received prior antibiotic treatment. When combined with microbial investigation, direct visualization of the tracheobronchial tree might be useful in determining the presence of bacterial pneumonia.