To conduct a retrospective multicenter study to assess the ability of a predictive algorithm to differentiate between children with Kingella kingae infection of the hip and those with transient synovitis.

Medical charts of 25 Israeli and 9 Spanish children aged 6-27 months with culture-proven K kingae arthritis of the hip were reviewed, and information on the 4 variables included in the commonly used Kocher prediction algorithm (body temperature, refusal to bear weight, leukocytosis, and erythrocyte sedimentation rate) was gathered.

Patients with K kingae arthritis usually presented with mildly abnormal clinical picture and normal serum levels of or near-normal acute-phase reactants. Data on all 4 variables were available for 28 (82%) children, of whom 1 child had none, 6 children had 1, 13 children had 2, 5 had 3, and only 3 children had 4 predictors, implying ≤40% probability of infectious arthritis in 20 (71%) children.

Because of the overlapping features of K kingae arthritis of the hip and transient synovitis in children younger than 3 years of age, Kocher predictive algorithm is not sensitive enough for differentiating between these 2 conditions. To exclude K kingae arthritis, blood cultures and nucleic acid amplification assay should be performed in young children presenting with irritation of the hip, even in the absence of fever, leukocytosis, or a high Kocher score.