Classification of Early Onset Scoliosis has Excellent Interobserver and Intraobserver Reliability

Background: Classification systems can be useful tools for clinical care and research but must be proven as reliable. The purpose of this study was to evaluate the interobserver and intraobserver reliability of the Classification of Early Onset Scoliosis (C-EOS) scheme. Methods: After IRB approval, 50 cases were drawn from a prospectively collected database of patients with EOS. Cases were selected using a stratified randomization scheme based on etiology. These cases were used to create an internet survey that was sent to pediatric orthopaedic faculty, research coordinators, and fellows involved in EOS care and research. Participants were asked to classify each case and were provided with a written C-EOS scheme that could be referenced while they completed the survey. Surveys were sent to participants twice, 3 weeks apart, to assess both intraobserver and interobserver reliability. Fleiss k and Cohen k were used to assess interobserver and intraobserver reliability, respectively. Results: There were 36 total participants, 29 who completed the survey twice (21 faculty, 13 research coordinators, and 2 fellows). Overall Fleiss k coefficient for interobserver reliability was excellent across the major categories of etiology (0.84), major curve (0.93), and kyphosis (0.96). Overall intraobserver reliability was excellent with Cohen k values for etiology (0.92), major curve (0.96), and kyphosis (0.98). Faculty members had excellent agreement for etiology (0.90), major curve (0.91), and kyphosis (0.96). Research coordinators had high levels of agreement for etiology (0.78), and excellent for major curve (0.95) and kyphosis (0.96). Intraobserver reliability was excellent across all major categories for all groups: faculty, research coordinators, and fellows. Conclusions: The study shows high levels of interobserver and intraobserver agreement of the C-EOS scheme. The C-EOS scheme can be used as a reliable tool for classifying EOS patients for clinical communication and research. Level of Evidence: Level II-diagnostic study.