Sputum Gram Stain for Bacterial Pathogen Diagnosis in Community-acquired Pneumonia: A Systematic Review and Bayesian Meta-analysis of Diagnostic Accuracy and Yield

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.

Keywords: community-acquired pneumonia; diagnosis; meta-analysis; sensitivity and specificity; sputum Gram stain.

Figure 1.

Preferred Reporting Items for Systematic Review and Meta-Analyses flow diagram. Abbreviation: CAP, community-acquired pneumonia.

Figure 2.

Risk of bias and concerns regarding applicability of included studies. Abbreviations: BTS, British Thoracic Society; ID, identification number; QUADAS-2, Quality Assessment of Diagnostic Accuracy Studies 2.

Figure 3.

Diagnostic accuracy of good-quality sputum specimens for the diagnosis of Streptococcus pneumoniae (A and B), Haemophilus influenzae (C and D), and other bacteria (E). Cross-hair receiver operating characteristic (ROC) plots (A, C, and E) show reported prior-point estimates (shown as circles), and confidence intervals (shown as extended lines). ROC plots and hierarchical summary ROC curves (B and D) show individual study posterior-point estimates (the size of each circle is proportional to the sample size for each study). The dashed elliptical boundary represents the 95% credible region for the summary estimates (closed diamond). The standard (black) and latent-class model analyses based on the conditional dependence model (blue) and the conditional independence model (red) are presented. E, Causative bacteria and their diagnostic criteria of visual assessment (in parentheses). Abbreviations: GNDC, gram-negative diplococci; GNRl, large-sized gram-negative rods; GNRs, small-sized gram-negative rods; GPC, gram-positive cocci in clusters.

Figure 4.

Summary sensitivity and specificity of sputum Gram stain in community-acquired pneumonia. Diamonds represent point estimates. Extending lines represent the 95% credible interval of each estimate. Abbreviations: CrI, credible interval; DIC, deviance information criteria; US, United States.

Figure 5.

Summary positive and negative likelihood ratios of sputum Gram stain in community-acquired pneumonia. Diamonds represent point estimates. Extending lines represent the 95% credible interval of each estimate. Bold (drawn at 10 and 0.1) and thin (drawn at 5 and 0.2) dashed vertical lines represent clinically important thresholds that would make large and moderate shifts, respectively, in probability. Abbreviations: CrI, credible interval; DIC, deviance information criteria; NLR, negative likelihood ratio; PLR, positive likelihood ratio; US, United States.

Figure 6.

Summary diagnostic yield of sputum Gram stain in community-acquired pneumonia. Closed squares represent reported point estimates. Extended lines represent the 95% confidence interval of each estimate. Diamonds depict the meta-analytic result, with the center of the diamond and dashed vertical line representing the summary estimate and the width of the diamond representing the 95% CrI. Abbreviations: CrI, credible interval; ID, identification number.