Comparison of direct examination of three types of bronchoscopy specimens used to diagnose nosocomial pneumonia.

Objective: To compare direct examination of bronchial aspirate and plugged telescopic catheter specimens (PTC) with infected cell counts in bronchoalveolar lavage (BAL) specimens for the diagnosis of nosocomial pneumonia. Design: Prospective study of critically ill patients. Setting: Intensive care unit in a university hospital. Patients: A total of 64 patients hospitalized for >48 hrs with suspected nosocomial pneumonia. Interventions: Fiberoptic bronchoscopy with bronchial aspirate and quantitative protected specimen brush, PTC, and BAL cultures. PTC and bronchial aspirate specimens were Gram-stained. BAL specimens for infected cell counts were examined as described previously in the literature. Measurements and Main Results: Nosocomial pneumonia was diagnosed by the medical staff based on all available clinical, radiologic, laboratory test, and microbiological data and on the course before and after appropriate therapy. A total of 71% of patients were ventilated, and 70.1% were receiving antibiotics. Nosocomial pneumonia was diagnosed in 54% of the cases. On direct examination, sensitivity (Se) and specificity (Sp) of bronchial aspirate specimens were Se, 82% and Sp, 60%; of BAL with 5% infected cells, Se, 56% and Sp, 100%; of BAL with 3% infected cells, Se, 74% and Sp, 96%; of PTC specimens, Se, 65% and Sp, 76%; and of PTC specimens plus BAL with 3% infected cells, Se, 83% and Sp, 78%. BAL with 3% infected cells was significantly better for predicting nosocomial pneumonia than direct examination of bronchial aspirate or PTC specimens (p = .0012). When the BAL showed 3% infected cells, neither direct examination of bronchial aspirate nor direct examination of PTC specimens was useful (p = .24 and p = .38, respectively). Combined use of direct examination of PTC specimens plus BAL with 3% infected cells markedly improved sensitivity. The total cost of each procedure was taken into account for the final evaluation. Conclusions: Our data suggest that BAL with 3% infected cells is currently the only test whose predictive value for nosocomial pneumonia is sufficiently high to be of use for guiding the initial choice of antimicrobial class while waiting for quantitative culture results.