The crucial differentiation between septic arthritis and transient synovitis of the hip in children can be difficult. In 1999, Kocher et al introduced four clinical predictors which were highly predictive (99.6%) of septic arthritis. These included fever (temperature > or = 38.5 degrees C), inability to bear weight, white blood-cell count > 12.0 x 10(9) cells/L and ESR > or = 40 mm/hr; CRP > or = 20 mg/L was later added as a fifth predictor. We retrospectively evaluated these predictors to differentiate septic arthritis from transient synovitis of the hip in children over a four-year period in a primary referral general hospital. When all five were positive, the predicted probability of septic arthritis in this study was only 59.9%, with fever being the best predictor. When applied to low-prevalence diseases, even highly specific tests yield a high number of false positives and the predictive value is thereby diminished. Clinical predictors should be applied with caution when assessing a child with an irritable hip, and a high index of suspicion, and close observation of patients at risk should be maintained.