We compared five classification systems for clavicle fractures. The aim of this study was to evaluate the prognostic value of each system. Over a two-year period we reviewed all new radiographs of the shoulder region and identified 487 clavicle fractures. Each radiograph was classified using five classification systems. We reviewed all subsequent X-rays and clinical records until the patient was discharged. We assessed each classification system's prognostic value in predicting delayed/non-union. Our data show that 79.3% of clavicle fractures occur in the middle third, 19.3% in the lateral third and 1.4% in the medial third. The overall prevalence of delayed/non-union was 7.3%, with 3.2% requiring operative management and 4.1% developing asymptomatic non-union. The incidence of non-union in the lateral third was 9.6%, but only 0.4% required operative management. Craig's classification had the greatest prognostic value for lateral third fractures, and Robinson's classification had the greatest prognostic value for middle third fractures. Fractures of the clavicle are common injuries but non-union is an uncommon occurrence. Non-union is more common in the lateral third, but we found these to be mostly asymptomatic. Middle third fractures are more likely to require operative fixation. Middle third fractures should be classified according to Robinson's classification system and lateral third fractures according to Craig's classification. We did not assess sufficient medial third fractures for the data to be significant.