RESUMO
BACKGROUND: This study assessed the intraobserver and interobserver reliability of a binary classification system using an easy-to-remember acronym (the HGLS system--based on the reappraisal of Codman's description by Hertel et al) and compared it with the AO and Neer systems. MATERIALS AND METHODS: Forty-seven proximal humeral fractures in 47 patients treated at the Department of Orthopaedics and Trauma, Royal Adelaide Hospital, Adelaide, Australia, were identified in the period from July 2007 until January 2008. Fractures of the proximal humerus were examined with anteroposterior, lateral, and axillary radiographs. Three independent reviewers classified the fractures using the AO, Neer, and HGLS systems. Reclassification of the same fractures was undertaken after a 6-month interval, and interobserver and intraobserver correlation, by use of the κ statistic, was calculated for all 3 classification systems. RESULTS: The mean age of patients was 64.5 years (range, 16-95 years). The interobserver correlations for the AO system (κ value, 0.47) and Neer system (κ value, 0.44) were graded as poor and were consistent with the values of previously published studies. The HGLS classification showed good interobserver agreement for all 3 examiners at the first interpretation (κ value, 0.73) and second interpretation (κ value, 0.61). Good intraobserver agreement after a 6-month period was also seen for the HGLS classification (κ values, 0.87-0.92) compared with the AO system (κ, 0.61-0.71) and Neer system (κ, 0.42-0.77). CONCLUSION: The HGLS system provided a more reliable description of fractures of the proximal humerus compared with the Neer and AO systems. Further studies are necessary to assess the validity of the HGLS system.
