Introduction: The limping child poses a diagnostic challenge. The purpose of this study was to create a clinically useful algorithm of presenting variables to allow the exclusion of ‘musculoskeletal sepsis’ as a differential diagnosis in the child presenting with a limp.

Materials & Methods: This study represents the data collected on all limping children admitted to our centre over a 3-year period. Analysis was based on 229 admissions. Comparison was made between the group with septic arthritis or osteomyelitis and the group without infection, using univariate analysis. With logistic regression analysis, a model consisting of three independent multivariate predictors was constructed, to exclude infection.

Results: Patients with septic arthritis or osteomyelitis differed significantly from patients without infection with regard to duration of symptoms, presence of constitutional symptoms, temperature, white cell count and erythrocyte sedimentation rate (ESR), (p-values < 0.05). Multivariate analysis demonstrated that the best model to describe our patient population was based on three variables: duration of symptoms between 1 and 5 days, temperature > 37.0°C and ESR > 35mm/hr. When all three variables were present, the predicted probability of musculoskeletal infection was 0.66. When none of the three were present, the predicted probability of infection was 0.01.

Discussion: Diagnosis of septic arthritis or osteomyelitis is especially difficult in the early phase and there is no single variable that can serve as a definitive test. The significance of constitutional symptoms and duration of symptoms on univariate analysis emphasises the importance of careful history taking. C-reactive protein, while considered for inclusion, was excluded due to its limited availability at our institution.

Conclusion: The multivariate model enables us to rule out musculoskeletal infection with 99% certainty in limping children with none of these three presenting variables.