Evaluation of inter- and intra-observer reliability of current classification systems for subtrochanteric femoral fractures.

INTRODUCTION: To evaluate the inter- and intra-observer reliability of the Russell Taylor, Seinsheimer and AO classification systems, and to investigate whether or not the experience of the surgeon had any effect on the classification.

PATIENTS AND METHODS: All the radiographs of 35 patients with subtrochanteric femur fracture were classified by 16 observers using the Russell Taylor, Seinsheimer and 31-AO classifications. Two groups of observers were formed of eight orthopedic surgeons, each with at least five-year experience and eight orthopedic assistants, from six different hospitals, who were invited to participate in the study. All the observers reviewed all the X-rays at this first evaluation. At 6 weeks after the first evaluation, the same radiographs were presented to each observer again in a random order and all the observers were requested to classify the fractures again. To evaluate the inter- and intra-observer reliability, the Fleiss kappa and Cohen's kappa values were used.

RESULTS: In the inter-observer reliability, the mean values of the two evaluations for the Russell Taylor classification were determined to be κ:0.724 (substantial) for the specialists and κ:0.722 (substantial) for the assistants. Using the Seinsheimer classification, the mean values were κ:0.691 (substantial) for the specialists and κ:0.629 (substantial) for the assistants, and for the AO classification, the mean values were κ:0.279 (fair) for the specialists and κ:0.291 (fair) for the assistants. In the intra-observer reliability, the median values for the Russell Taylor classification were determined to be κ:0.955 (almost perfect) for the specialists and κ:0.855 (almost perfect) for the assistants. Using the Seinsheimer classification, the median values were κ:0.915 (almost perfect) for the specialists and κ:0.900 (almost perfect) for the assistants, and for the AO classification, the median values were κ:0.665 (substantial) for the specialists and κ:0.695 (substantial) for the assistants.

CONCLUSIONS: As both the Russell Taylor and Seinsheimer classifications were found to be more reliable and reproducible than the AO classification for subtrochanteric femoral fractures, they can be considered to be more valuable in clinical practice and communication. The experience of the surgeons was not found to have any significance in the evaluation of these three classification systems in these types of fractures.