ABSTRACT
BACKGROUND: Although the Neer classification is widely used for the assessment of proximal humeral fractures, its reproducibility has been challenged. The purpose of this study was to evaluate the reproducibility of the conventional Neer classification and a modified classification that defined fracture displacement with respect to the humeral head fragment. METHODS: The fracture patterns in 80 cases of proximal humeral fractures were independently assessed by 6 observers. The cases were grouped according to the conventional Neer classification using radiographs followed by computed tomography (CT) scans by each examiner twice with a 1-month interval. The fractures were then classified with the modified Neer classification, which defined displacement of the fragment as separation of more than 1 cm or angulation of more than 45° from the humeral head fragment, twice with a 1-month interval. Kappa coefficients of the conventional and modified Neer classifications were compared. RESULTS: The modified classification showed significantly higher intra-observer agreement than the conventional classification, both for radiographs (P = .028) and for CT scans (P = .043). Intra-observer agreement was also significantly higher for the modified classification than for the conventional classification, both for radiographs (P = .001) and for CT scans (P < .001). CONCLUSIONS: The present study showed that agreement for the Neer classification could be improved when fracture displacement was defined as separation or angulation from the humeral head. Considering vascularity to the humeral head, furthermore, the modified method might be more helpful for predicting patients' prognosis than the conventional Neer classification.