Introduction: Septic arthritis presenting as an acutely inflamed knee (AIK) requires prompt diagnosis and treatment to prevent irreversible joint damage and systemic complications. A lack of high-quality diagnostic criteria hinders decision making. A multivariable prediction model developed among nonveterans demonstrated strong performance characteristics. This was converted into the Septic Arthritis Risk Calculator (SARC), providing personalized risk estimates for native knee septic arthritis. The purpose of this study was to externally validate the accuracy and calibration of SARC. Methods: A retrospective cohort study was conducted for veterans with a native AIK at a Veterans Affairs Medical Center from 2007 to 2022. Demographics, comorbidities, history, physical examination, laboratory and radiographic findings, and confirmation of septic arthritis were collected. Septic arthritis risk was calculated. Discrimination was assessed by receiver operating characteristic analysis. Calibration was assessed via Hosmer-Lemeshow goodness-of-fit and Cox calibration regression (statistical significance for P < 0.05). Results: The cohort consisted of 241 veterans with 50 cases (20.7%) of septic arthritis. The area under the curve was high (0.89) with 90% of cases correctly classified. The model fit the cohort well, supported by Hosmer-Lemeshow goodness-of-fit ( P = 0.43) and Cox calibration regression yielding an estimated slope of 1.0 (95% CI: 0.74 to 1.3) and intercept near zero (−0.51 to 0.51). Conclusion: SARC was accurate and well calibrated among veterans as in the original nonveteran cohort. We recommend its use in the setting of an AIK. Further research will involve implementation through an online decision support tool for point of care clinical use.
Kunz et al. (Sun,) studied this question.