Clinical prediction model performance in differentiating septic arthritis from transient synovitis: A multi-center study

Objective: Differentiating septic arthritis from transient synovitis in children is challenging. This study aimed to determine the diagnostic value for distinguishing these two conditions and to develop an effective clinical prediction model based on multi-center clinical data to identify and differentiate between the two diseases more comprehensively and accurately. Methods: We retrospectively analyzed data of children aged under 18 years who were hospitalized in eight specialized children's hospitals in China from 2013 to 2021. The data included age, sex, initial hospital admission temperature, history of prodromal respiratory infection, prodromal history of strenuous exercise, etc. Laboratory test results, including that of erythrocyte sedimentation rate (ESR), serum white blood cell count (WBC), C-reactive protein (CRP), platelet count (PC), pathogenic bacterial count and other information. The common sites of bone and joint infection in children are the hip and knee joints. Therefore, to ensure the prediction model's reliability, we established three clinical prediction models: one for septic hip arthritis and transient hip synovitis, one for septic knee arthritis and transient knee synovitis, and one for septic arthritis and transient synovitis in either joint which we merged all the data of the hip and knee joints into one data set to build a clinical prediction model. In addition, we conducted single-factor and multi-factor analyses on the data classified by these models. Results: This study collected data of 819 children from 8 tertiary children's hospitals, including 265 patients with septic arthritis (mean age 5.55±7.16 years) and 554 patients with transient synovitis (mean age 4.8±3.2 years). In 265 blood and joint aspirate cultures of septic arthritis, the most common causative organism was Staphylococcus aureus (59%). To ensure the quality and reliability of the prediction models, we established three clinical prediction models. For septic hip arthritis, a retrospective study based on six clinical predictors was a history of prodromal respiratory tract infection (HRTI), temperature>37.5 ºC, ESR>20 mm/h, CRP>10 mg/L, red blood cell distribution width (RDW)>50%, and WBC>11×109 /L. When these six factors were present, the probability of septic hip arthritis was 99.99%. For septic knee arthritis, a retrospective study based on three clinical predictors, the predictors were ESR>20 mm/h, CRP>10 mg/L, and absolute monocyte count (AMONO)>0.74×109/L. When these three factors were present, the probability of having septic knee arthritis was 94.68%. For septic arthritis (septic hip arthritis or septic knee arthritis), a retrospective study based on six clinical predictors, the predictors were male children, history of HRTI), temperature>37.5 ºC, ESR>20 mm/hr, PC > 407×10 ⁹ /L and CRP>10 mg/L. When these six factors were present, the probability of septic arthritis was 99.65%. Conclusion: This study used multi-center clinical data to construct a new clinical prediction model for children with septic arthritis. The clinical prediction model algorithm developed here outperformed that of previous studies in the probability of diagnosing septic arthritis. In addition to considering body temperature, WBC, ESR, and CRP, we identified new clinical predictors such as sex, history of HRTI, RDW, PC and AMONO.