Inter- and intraobserver agreement of three classification systems for lateral clavicle fractures - reliability comparison between two specialist groups

Background Although of great value in the management of lateral clavicle fractures, substantial variation in their classification exists. We performed a retrospective study to address the inter- and intraobserver reliability of three different classification systems for lateral clavicle fractures. Methods Radiographs of 20 lateral clavicle fractures that represented a full spectrum of adult fracture patterns were graded by five experienced radiologists and five experienced trauma surgeons according to the Orthopaedic Trauma Association (OTA), the Neer, and the Jager/Breitner classification systems. This evaluation was performed at two different time points separated by 3 months. To measure the observer agreement, the Fleiss kappa coefficient (kappa) was applied and assessed according to the grading of Landis and Koch. Results The overall interobserver reliability showed a fair agreement in all three classification systems. For the OTA classification system, the interobserver agreement showed a mean kappa value of 0.338 ranging from 0.350 (radiologists) to 0.374 (trauma surgeons). Kappa values of the interobserver agreement for the Neer classification system ranged from 0.238 (trauma surgeons) to 0.276 (radiologists) with a mean kappa of 0.278. The Jager/Breitner classification system demonstrated a mean kappa value of 0.330 ranging from 0.306 (trauma surgeons) to 0.382 (radiologists). The overall intraobserver reliability was moderate for the OTA and the Jager/Breitner classification systems, while the overall intraobserver reliability for the Neer classification system was fair. The kappa values of the intraobserver agreements showed, in all classification systems, a wide range with the OTA classification system ranging from 0.086 to 0.634, the Neer classification system ranging from 0.137 to 0.448, and a range from 0.154 to 0.625 of the Jager/Breitner classification system. Conclusions The low inter- and intraobserver agreement levels exhibited in all three classification systems by both specialist groups suggest that the tested lateral clavicle fracture classification systems are unreliable and, therefore, of limited value. We should recognize there is considerable inconsistency in how physicians classify lateral clavicle fractures and therefore any conclusions based on these classifications should be recognized as being somewhat subjective.