Septic arthritis score (SAS) - a novel clinical prediction model for the probability of septic arthritis in the adult native knee.

Background: Patients presenting with an acutely painful swollen joint represent a diagnostic challenge. We aimed to develop a clinical prediction model for septic arthritis (SA) in the adult native knee.

Methods: We screened all synovial cultures in south Sweden in 2020 and 2021. We included cultures taken in the emergency department from adults' native knees where SA was considered a differential diagnosis based on medical chart review. We developed a prediction model using logistic regression and performed internal validation using bootstrapping. We present a nomogram and an online calculator ( http://sascore.org ) for individual risk estimation, net benefit compared to usual care and treatment threshold recommendations.

Results: A total of 668 patients were included from 2996 screened synovial cultures. The final septic arthritis score (SAS) included four variables: synovial-to-serum glucose quotient, synovial white blood cell count, abnormal synovial fluid appearance on visual inspection, and triage priority according to Rapid Emergency Triage and Treatment System (RETTS) vital signs. SAS had an optimism-adjusted area under the receiver operating characteristics curve of 0.971 (95% bootstrap confidence interval: 0.957 to 0.987). Clinicians provided empirical intravenous antibiotics to 47 out of 51 patients with a final diagnosis of SA and to 244 out of 617 patients without SA (92% sensitivity, 60% specificity). SAS had 92% sensitivity and 92% specificity at 10% probability for SA treatment threshold and 100% sensitivity and 79% specificity at 2% treatment threshold.

Conclusion: The use of SAS would theoretically avoid 50-82% of unnecessary empirical antibiotics as compared to usual care in our cohort with retained or improved identification of actual septic arthritis of the native knee. External validation is warranted before clinical use.