Abstract
BACKGROUND: The clinical role of sputum Gram stain (SGS) in community-acquired pneumonia (CAP) diagnosis remains controversial. A 1996 meta-analysis of the diagnostic accuracy of SGS reported heterogeneous results. To update the available evidence, we performed a systematic review and a Bayesian standard and latent-class model meta-analysis. METHODS: We searched Medline, Embase, and Cochrane Central by 23 August 2018 to identify studies reporting on the diagnostic accuracy, yield (percentage of patients with any pathogen[s] correctly identified by SGS), and clinical outcomes of SGS in adult patients with CAP. Two reviewers extracted the data. We quantitatively synthesized the diagnostic accuracy and yield, and descriptively analyzed other outcomes. RESULTS: Twenty-four studies with 4533 patients were included. The methodological and reporting quality of the included studies was limited. When good-quality sputum specimens were selected, SGS had a summary sensitivity of 0.69 (95% credible interval [CrI], .56-.80) and specificity of 0.91 (CrI, .83-.96) for detecting Streptococcus pneumoniae, and a sensitivity of 0.76 (CrI, .60-.87) and specificity of 0.97 (CrI, .91-.99) for Haemophilus influenzae. Adjusted analyses accounting for imperfect reference standards provided higher-specificity estimates than the unadjusted analyses. Bacterial pathogens were identified in 73% (CrI, 26%-96%) of good-quality specimens, and 36% (CrI, 22%-53%) of all specimens regardless of quality. Evidence on other bacteria was sparse. CONCLUSIONS: SGS was highly specific to diagnose S. pneumoniae and H. influenzae infections in patients with CAP. With good-quality specimens, SGS can provide clinically actionable information for pathogen-directed antibiotic therapies.