Evaluation for septic arthritis of the native adult knee is aided by multivariable assessment.

ABSTRACT

INTRODUCTION: Evaluation of suspected septic arthritis of the native adult knee is a common diagnostic dilemma. Pre-aspirate criteria predictive of septic arthritis do not exist for the adult knee and investigations of aspiration results (cell count, differential, gram stain and crystal analysis) have been limited to univariate analyses. Given numerous clinical variables inform the risk of septic arthritis, multivariable analysis that incorporates all clinically available information is critical to allowing accurate decision-making. METHODS: We retrospectively identified 455 cases of potential septic arthritis of a native adult knee at a tertiary health system from 2012 to 2017, of which 281 underwent aspiration. We recorded demographics, comorbidities, history, exam, laboratory, and radiographic data. Among aspirated cases, we performed univariate analyses of all variables for association with septic arthritis followed by multivariable logistic regression analysis. RESULTS: Septic arthritis was confirmed in 61 of 281 patients who underwent aspiration. Independent associations of risk for septic arthritis included synovial fluid WBC ≥ 30,000 (Odds Ratio 90.8, 95% Confidence Interval 26.6-310.1, p < 0.001), bacteria reported on synovial fluid gram stain (OR 21.5, 95% CI 3.9-119.2, p < 0.001), duration of pain >2 days (OR 6.9, 95% CI. 2.3-20.9, p < 0.001), history of septic arthritis at any joint (OR 5.0, 95% CI 1.1-23.4, p = 0.039), clinical effusion (OR 4.8, 95% CI 1.2-20.0, p = 0.030). Independent associations protective against septic arthritis included presence of synovial fluid crystals (OR 0.1, 95% CI 0.1-0.4, p < 0.001). The multivariable model was highly accurate in discriminating between septic and aseptic cases (AUC = 0.942). A web-based tool was created to aid clinical decision-making. CONCLUSION: When evaluating for septic arthritis of a native adult knee, several independent associations were identified for variables related and unrelated to joint aspiration. The associated multivariable model discriminated very well between patients with and without septic arthritis, outperforming previous univariate assessments. A web-based tool was created that estimates the probability of septic arthritis based on this model. This may aid decision-making in complex clinical scenarios.