THE NEER-CLASSIFICATION-SYSTEM FOR PROXIMAL HUMERAL FRACTURES - AN ASSESSMENT OF INTEROBSERVER RELIABILITY AND INTRAOBSERVER REPRODUCIBILITY

The radiographs of fifty fractures of the proximal part of the humerus were used to assess the interobserver reliability and intraobserver r eproducibility of the Neer classification system. A trauma series cons isting of scapular anteroposterior, scapular lateral, and axillary rad iographs was available for each fracture. The radiographs were reviewe d by an orthopaedic shoulder specialist, an orthopaedic traumatologist , a skeletal radiologist, and two orthopaedic residents, in their fift h and second years of postgraduate training. The radiographs were revi ewed on two different occasions, six months apart. Interobserver relia bility was assessed by comparison of the fracture classifications dete rmined by the five observers. Intraobserver reproducibility was evalua ted by comparison of the classifications determined by each observer o n the first and second viewings. Kappa (kappa) reliability coefficient s were used. All five observers agreed on the final classification for 32 and 30 per cent of the fractures on the first and second viewings, respectively. Paired comparisons between the five observers showed a mean reliability coefficient of 0.48 (range, 0.43 to 0.58) for the fir st viewing and 0.52 (range, 0.37 to 0.62) for the second viewing. The attending physicians obtained a slightly higher kappa value than the o rthopaedic residents (0.52 compared with 0.48). Reproducibility ranged from 0.83 (the shoulder specialist) to 0.50 (the skeletal radiologist ),,vith a mean of 0.66. Simplification of the Neer classification syst em, from sixteen categories to six more general categories based on fr acture type, did not significantly improve either interobserver reliab ility or intraobserver reproducibility.