Objective: To evaluate the preparation and interpretation of sputum Gram stains by housestaff physicians in the assessment of patients with community-acquired pneumonia.
Design: A prospective, multicenter study.
Setting: Two university-affiliated hospitals in Pittsburgh.
Patients: Ninety-nine cases of clinically and radiographically established pneumonia occurring in 97 patients. Diagnostic test assessment: Housestaff and microbiology personnel prepared a Gram stain for each case of pneumonia. Housestaff assessed the presence and identity of a predominant microbial organism on the slides they prepared. Two senior staff microbiologists, blinded to patient and preparer, evaluated all slides for preparation, sputum purulence, and identification of the predominant organism. Two reference standards were used to assess the sensitivity, specificity, and predictive values of housestaff's Gram-stain interpretations: 1) senior staff microbiologists' determinations of the microbes present using the slides without benefit of culture results, and 2) the etiologic agent derived from results of sputum culture, blood culture, or serology.
Measurements and main results: Housestaff physicians completed a Gram stain in 58% of the pneumonia episodes. Gram stains were not made in 42% of cases, primarily because patients were unable to produce sputum. Fifteen percent of housestaff's smears were judged inadequately prepared, compared with 3% for the laboratory personnel (p less than 0.01). Housestaff obtained purulent sputum samples significantly more often than did nursing personnel (58% versus 38%; p less than 0.01). Housestaff's Gram stains were 90% sensitive for detecting pneumococcus, with a 50% false-positive rate. The sensitivity of the Gram stain was less for identification of Haemophilus influenzae than for identification of Streptococcus pneumoniae. A single antimicrobial agent was chosen as initial therapy for 50% of the patients in whom housestaff identified a predominant organism, compared with 30% in whom a predominant organism was not identified (p less than or equal to 0.05).
Conclusions: Although housestaff obtained purulent sputum samples more frequently than did nursing personnel, they made systematic errors in the preparation and interpretation of Gram-stained slides. Housestaff physicians should receive formal training in the preparation and interpretation of Gram stains; the specific defects elucidated in this study warrant special attention.