Classification of thoracic and lumbar injuries: An analysis of interobserver reliability

The purpose of a fracture classification is to help the surgeon to choose an appropriate method of treatment for each and every fracture occuring in a particular anatomical region. The classification tool should not only suggest a method of treatment, it should also provide the surgeon with a reasonably precise estimation of the outcome of that treatment. But to use a classification before its workability has been proved is inapproproiate and can lead to confusion and more conflicting results. Any classification system should be proved to be a workable tool before it is used in a discriminatory or predictive manner. The radiographs of fourteen fractures of the lumbar spine were used to assess the interobserver reliability of the AO classification system. The radiographs and CT scans were reviewed in twenty two hospitals experienced with spinal trauma. The mean interobserver agreement for all fourteen cases was found to be 67 % (41-91 %),when only the three main types (A, B, C) were used. The corresponding kappa value of the interobserver reliability showed a coefficient of 0,33 (range, 0,30 to 0,35). The reliability decreased by increasing the categories. For some injuries the interobserver reliability was found to be over 90 % and also for the recommended therapeutic procedure there was an acceptable agreement. But the decision between an posterior approach alone or an additionally anterior procedure seems to be the most important question in treatment of spinal injuries at that time.