The interobserver reliability of classification systems for radial head fractures: the Hotchkiss modification of the...

BACKGROUND: Radial head fractures are common injuries, and there is little information on the reliability of classification systems for such injuries. The purpose of our study was to report the interobserver reliability of 2 commonly used classification systems: the Hotchkiss modification of the Mason classification and the AO classification systems. METHODS: We compiled the radiographs from a cohort series of 43 patients with radial head fractures, and 5 observers classified the radiographs according to both classification systems. Additionally, we collapsed the systems, with types II and III combined for the Hotchkiss classification and the final digit dropped for the AO classification. We calculated percent agreement, the kappa statistic and the associated 95% confidence intervals (CIs). RESULTS: The mean percent agreement was 72.3% (95% CI 65.8%-78.9%) for the Hotchkiss classification and 37.7% (95% CI 30.5%-44.9%) for the AO classification. The kappa statistic was 0.585 (0.541-0.661) for the Hotchkiss classification and 0.261 (0.240-0.350) for the AO classification. The mean percent agreement was 89.3% (86.6%-92.0%) for the consolidated Hotchkiss classification and 67.4% (54.6%-80.3%) for the consolidated AO classification. The kappa statistic was 0.760 (0.691-0.805) for the consolidated Hotchkiss classification and 0.455 (0.372-0.521) for the consolidated AO classification. CONCLUSION: The interobserver reliability for the Hotchkiss modification of the Mason classification was moderate, and that for the AO classification was fair according to the criteria of Landis and Koch. Collapsing the Hotchkiss classification improved the reliability to substantial, and collapsing the AO system improved reliability to the lower end of moderate.