Inter- and intraobserver reliability assessment of the 2018 AO/OTA classification for high-energy pelvic ring injuries: A retrospective study

Background. - A fracture classification system should be a reliable and reproducible means of communication between different observers. It should be logical, comprehensible, and shouldn't contain an unmanageable number of categories. The aim of this study was to assess the intra- and interobserver agreement and reliability of the revised 2018 AO/OTA classification for high-energy pelvic ring injuries (PRI), at the level of the types, groups, subgroups and qualifications. Hypothesis. - Agreement and reliability of the revised 2018 AO/OTA classification for high-energy PRI are improved when compared to previous versions of the classification. Patients and methods. - Plain radiographs and computed tomography images of a consecutive series of 86 adult patients admitted at a level I trauma center with a high-energy PRI between 01.01.2014 and 31.12.2016 were retrospectively analyzed. Three orthopedic surgeons independently classified these PRI using the 2018 AO/OTA and the Young and Burgess classifications. The senior surgeon analyzed all injuries twice, at 6 months interval, to determine intraobserver reliability. Classification agreement was assessed using percent agreement and classification reliability was assessed using kappa coefficients. Results. - For the intraobserver analysis, injury classifications with the 2018 AO/OTA classification were concordant in 88% of cases (type), 74% (group), 66% (subgroup) and 49% (qualification). Respective kappa coefficients were 0.79, 0.68, 0.62 and 0.47. Interobserver agreement declined from 77% (type) to 42% (group), 36% (subgroup) and 24% (qualification). Respective kappa coefficients were 0.72, 0.48, 0.48 and 0.37. Intraobserver (respectively interobserver) percent agreement with the Young and Burgess classification was 76% (50%) and kappa coefficient was 0.69 (0.51). Discussion. - The 2018 AO/OTA classification is a reliable tool for daily clinical use and for research purpose at the fracture type level but not at the group, subgroup and qualification levels. These results compare favorably with previously published data for older versions of the classification and may represent an improvement of the AO/OTA classification system in terms of reliability. (C) 2021 L'Auteur(s). Publie par Elsevier Masson SAS.