Inter-observer reliability of alternative diagnostic methods for proximal humerus fractures: a comparison between attending surgeons and orthopedic residents in training

BACKGROUND: Proximal humerus fractures are frequent, and several studies show low diagnostic agreement among the observers, as well as an inaccurate classification of these lesions. The divergences are generally correlated with the experience of the surgeons as well as the diagnostic methods used. This paper challenges these problems including alternative diagnostic methods such as 3D models and augmented reality (holography) and including the observers’ period of medical experience as a factor. METHODS: Twenty orthopedists (ten experts in shoulder surgery and ten experts in traumatology) and thirty resident physicians in orthopedics classified nine proximal humerus fractures randomly distributed as x-ray, tomographies, 3D models and holography, using AO/ASIF and Neer’s classification. In the end, we evaluated the intra- and inter-observer agreement between diagnostic methods and whether the experience of the observers interfered in the evaluations and the classifications used. RESULTS: We found overall kappa coefficients ranging from 0.241 (fair) to 0.624 (substantial) between the two classifications (AO / ASIF and Neer), concerning the diagnostic methods used. We identified image modality differences (p = 0.017), where 3D models presented an average kappa coefficient value superior to that of tomographies. There were no differences between kappa scores for x-ray and holography compared to the others. The kappa scores for AO / ASIF classification and Neer classification and subdivided by observer period of experience showed no differences concerning the diagnostic method used. CONCLUSIONS: 3D models can substantially improve diagnostic agreement for proximal humerus fractures evaluation among experts or resident physicians. The holography showed good agreement between the experts and can be a similar option to x-ray and tomography in the evaluation and classification of these fractures. The observers’ period of experience did not improve the diagnostic agreement between the image modalities studied. TRIAL REGISTRATION: Registered in the Brazil Platform under no. CAAE 88912318.1.0000.5487.