Context In patients who present with an acutely painful and swollen joint, prompt identification and treatment of septic arthritis can substantially reduce morbidity and mortality.
Objective To review the accuracy and precision of the clinical evaluation for the diagnosis of nongonococcal bacterial arthritis.
Data Sources Structured PubMed and EMBASE searches (1966 through January 2007), limited to human, English-language articles and using the following Medical Subject Headings terms: arthritis, infectious, physical examination, medical history taking, diagnostic tests, and sensitivity and specificity.
Study Selection Studies were included if they contained original data on the accuracy or precision of historical items, physical examination, serum, or synovial fluid laboratory data for diagnosing septic arthritis.
Data Extraction Three authors independently abstracted data from the included studies.
Data Synthesis Fourteen studies involving 6242 patients, of whom 653 met the gold standard for the diagnosis of septic arthritis, satisfied all inclusion criteria. Two studies examined risk factors and found that age, diabetes mellitus, rheumatoid arthritis, joint surgery, hip or knee prosthesis, skin infection, and human immunodeficiency virus type 1 infection significantly increase the probability of septic arthritis. Joint pain (sensitivity, 85%; 95% confidence interval [CI], 78%-90%), a history of joint swelling (sensitivity, 78%; 95% CI, 71%-85%), and fever (sensitivity, 57%; 95% CI, 52%-62%) are the only findings that occur in more than 50% of patients. Sweats (sensitivity, 27%; 95% CI, 20%-34%) and rigors (sensitivity, 19%; 95% CI, 15%-24%) are less common findings in septic arthritis. Of all laboratory findings readily available to the clinician, the 2 most powerful were the synovial fluid white blood cell (WBC) count and percentage of polymorphonuclear cells from arthrocentesis. The summary likelihood ratio (LR) increased as the synovial fluid WBC count increased (for counts <25 000/μL: LR, 0.32; 95% CI, 0.23-0.43; for counts ≥25 000/μL: LR, 2.9; 95% CI, 2.5-3.4; for counts >50 000/μL: LR, 7.7; 95% CI, 5.7-11.0; and for counts >100 000/μL: LR, 28.0; 95% CI, 12.0-66.0). On the same synovial fluid sample, a polymorphonuclear cell count of at least 90% suggests septic arthritis with an LR of 3.4 (95% CI, 2.8-4.2), while a polymorphonuclear cell count of less than 90% lowers the likelihood (LR, 0.34; 95% CI, 0.25-0.47).
Conclusions Clinical findings identify patients with peripheral, monoarticular arthritis who might have septic arthritis. However, the synovial WBC and percentage of polymorphonuclear cells from arthrocentesis are required to assess the likelihood of septic arthritis before the Gram stain and culture test results are known.