Purpose: Classification systems for orthopaedic injuries are used in clinical care and research to allow for communication about the diagnosis, prognosis, treatment, and outcome of fractures and dislocations. The inter-observer reliability of the Hotchkiss modification of the Mason classification system and the AO classification system were evaluated to assess their reliability in classifying radial head fractures.

Methods: Forty-three consecutive fractures with and without other associated injuries were identified from the practice of a single sub-specialist orthopaedic surgeon. Their radiographs were compiled and all identifying marks were hidden. Five observers classified the radiographs according to each classification system. Percent agreement was calculated pair-wise between observers for each system. The mean percent agreement and 95% confidence intervals were calculated for each system. Additionally, the systems were collapsed, with types II and III combined for the Hotchkiss system, and the final digit dropped for the AO system. The mean percent agreement and 95% confidence intervals were then recalculated.

Results: The percent agreement for the Hotchkiss modification of the Mason classification system was 72.3% (95% CI 65.8% to 78.9%). The percent agreement for the AO classification system was 37.7% (95% CI 30.5% to 44.9%). The percent agreement for the collapsed Hotchkiss system was 89.3% (95% CI 86.6% to 92.0%). The percent agreement for the collapsed AO system was 67.4% (95% CI 54.6% to 80.3%).

Conclusions: The inter-observer reliability for the AO classification system had a low percent agreement, possibly reflecting the complexity of the system. Collapsing the AO system improved the percent agreement, however it still could be considered fair. The inter-observer reliability for the Hotchkiss modification of the Mason classification system was higher. However, when the lower end of the 95% confidence interval is taken into consideration, the reliability of the system could be considered fair. Collapsing this system did improve the percent agreement into what could be considered a good range, suggesting that this system may be able to reliably identify fractures requiring operative treatment.

Funding : Other Education Grant

Funding Parties : Alberta Heritage Foundation for Medical Research