Interobserver Reliability of 2 Classification Systems for Open Hand Fractures

Purpose: The aim of the study is to determine the reliability of the Tulipan-Ilyas (TI) classification among hand surgeons by utilizing several case studies and to test the null hypothesis that the TI system would have superior interobserver reliability than the Gustilo-Anderson (GA) system for classifying and guiding treatment for open fractures of the hand. Methods: Institutional review board exemption for this study was confirmed. All surgeons were unaware and blinded to the development and origin of the newly proposed TI classification system. Interobserver reliability was computed using the multi-rater κ described by Fleiss. The κ values were interpreted according to the guidelines of Landis and Koch. Interrater reliability was computed utilizing methods described by Shrout and Fleiss. Results: The overall κ value for interobserver reliability was 0.339 (95% confidence interval [CI], 0.304-0.376) for the GA classification and 0.443 (95% CI, 0.423-0.463) for the TI classification, indicating fair and moderate agreement, respectively. According to the intraclass correlation coefficients, the interrater agreement for average measures for both classification systems was excellent, with intraclass correlation coefficient (ICC) values, including the lower CIs, all greater than 0.91. However, for the ICCs for single measures, generalizing only to one rater, the interrater agreement for the GA classification (0.595, 95% CI, 0.387-0.836) was only good, but for the TI classification system was excellent (0.958, 95% CI, 0.912-0.987). Conclusion: The authors recommend the TI classification instead of the GA classification system for use in open fractures of the hand because of its superior interobserver reliability and its hand-specific nature.