Multi-centre evaluation of Gram stain in the diagnosis of septic arthritis.

Introduction: Gram stain of synovial fluid is a rapid test for the diagnosis of native joint septic arthritis. Single-centre studies have suggested Gram stain will miss a considerable proportion of patients who are subsequently synovial-fluid-culture-positive or polymerase chain reaction (PCR)-positive. The object of this study was to reassess Gram stain in a large, multi-centre cohort of patients from the United Kingdom (UK) and Ireland. Methods: The study was a retrospective analysis combining two large datasets. We defined septic arthritis microbiologically as at least one positive joint aspirate culture and/or PCR test. "Best case" and "worst case" definitions were applied depending on the likelihood organisms were true infecting pathogens. Results: Gram stain missed a high proportion of culture-/PCR-positive patients using both the best (74 % missed) and worst (81 % missed) case definitions. Using the best case definition, the sensitivity of Gram stain was 0.26, specificity 0.99, positive predictive value 0.84, negative predictive value 0.87, accuracy 0.87, and area under the receiver operator curve 0.62 (95 % CI 0.57 to 0.68, $p<0.001$ ). False positive Gram stains were infrequent (1 %). Age, joint involved, and other synovial fluid characteristics were less predictive of a positive culture/PCR than Gram stain. Conclusions: While a positive synovial fluid Gram stain should always be considered to indicate potential septic arthritis, a negative Gram stain, regardless of synovial fluid crystals or white cell count, should not be used to rule out septic arthritis. The value of Gram stain as an urgent out-of-hours test for septic arthritis is open to considerable debate.