To develop a clinical prediction rule that could accurately identify children with active rickets in countries where nutritional rickets is common, we prospectively recorded clinical features in 736 Nigerian children aged 18 months and older presenting with leg deformities or inability to walk. We scored radiographs of the wrists and knees for active rickets of the growth plates. Sensitivities and specificities of clinical variables for radiographically active rickets were calculated and, using logistic regression, we derived a clinical prediction rule. The prediction rule was tested in a validation set of 89 children. Wrists and costochondral enlargement were the clinical signs with the best combination of sensitivity (72% and 76%, respectively) and specificity (81% and 64%, respectively) for active rickets. Age < 5 years, height-for-age Z-score < -2, leg pain during walking, wrist enlargement and costochondral enlargement were independently predictive of active rickets (p < 0.01 for each in multivariate model). In the validation set, any three of these clinical features accurately identified 87% of children with active rickets, whereas only 24% of those without active rickets had three or more features. We conclude that clinical features can be used to identify children with active rickets.