Intraobserver and Interobserver Agreement in the Measurement of Displaced Humeral Medial Epicondyle Fractures in Children

Background: Fractures of the humeral medial epicondyle occur frequently in children. The decision to pursue operative or nonoperative treatment often hinges on the amount of perceived fracture displacement. This study was performed to assess both intraobserver and interobserver agreement in the measurements of displacement of these fractures on radiographs by orthopaedic surgeons with various levels of training. Methods: We performed a retrospective review of the radiographs of thirty-eight patients with a fracture of the medial epicondyle of the humerus. Digital anteroposterior, lateral, and oblique radiographs of each involved elbow made at presentation were presented to five separate reviewers with different levels of orthopaedic training, including two junior residents (junior residents 1 and 2), one fellow, one junior attending surgeon, and one senior attending surgeon. Each reviewer recorded the amount of perceived displacement in millimeters. A difference of >2 mm between measurements represented clinical disagreement between reviewers. Intraobserver and interobserver agreement was assessed by calculating both the intraclass correlation coefficient and the percentage of clinical disagreement between ratings. Results: The intraclass correlation coefficients for intraobserver agreement regarding the measurements on the anteroposterior radiographs were 0.24 (95% confidence interval, 0.00 to 0.68) for junior resident 1, 0.82 (95% confidence interval, 0.41 to 0.95) for junior resident 2, 0.83 (95% confidence interval, 0.46 to 0.96) for the senior attending surgeon, 0.92 (95% confidence interval, 0.69 to 0.98) for the junior attending surgeon, and 0.98 (95% confidence interval, 0.92 to 1.00) for the fellow. The combined intraclass correlation coefficient for intraobserver agreement was 0.76. The reviewers as a group disagreed with their own measurements an average of 26% of the time. The intraclass correlation coefficient for interobserver reliability with regard to the measurements on the anteroposterior radiographs for the group was 0.80 (95% confidence interval, 0.64 to 0.89), and the reviewers disagreed with each other an average of 54% of the time. The intraclass correlation coefficient for interobserver agreement was 0.28 (95% confidence interval, 0.03 to 0.76) for the measurements on the lateral radiographs and 0.62 (95% confidence interval, 0.34 to 0.89) for the measurements on the oblique radiographs, with reviewers disagreeing an average of 87% of the time with regard to the measurements on the lateral radiographs and 64% of the time with regard to the measurements on the oblique radiographs. Conclusions: Intraobserver agreement with regard to measurement of displacement of medial epicondyle fractures of the humerus varied among the reviewers but was low overall. Interobserver agreement was best for the measurements on the anteroposterior radiographs, but this was also low overall. These findings cast doubt on whether the amount of perceived displacement should be used as a criterion for choosing operative or nonoperative management of fractures of the humeral medial epicondyle. Agreement may be improved to acceptable levels by adopting a standard set of measurement guidelines, which include use of the anteroposterior radiograph when possible and consistently measuring at the point of maximal displacement.