Classification system for cervical spine deformity morphology: a validation study

Abstract

OBJECTIVE The objective of this study was to initially validate a recent morphological classification of cervical spine deformity pathology. METHODS The records of 10 patients for each of the 3 classification subgroups (flat neck, focal deformity, and cervicothoracic), as well as for 8 patients with coronal deformity only, were extracted from a prospective multicenter database of patients with cervical deformity (CD). A panel of 15 physicians of various training and professional levels (i.e., residents, fellows, and surgeons) categorized each patient into one of the 4 groups. The Fleiss kappa coefficient was utilized to evaluate intra- and interrater reliability. Accuracy, defined as properly selecting the main driver of deformity, was reported overall, by morphotype, and by reviewer experience. RESULTS The overall classification demonstrated a moderate to substantial agreement (round 1: interrater Fleiss kappa = 0.563, 95% CI 0.559–0.568; round 2: interrater Fleiss kappa = 0.612, 95% CI 0.606–0.619). Stratification by level of training demonstrated similar mean interrater coefficients (residents 0.547, fellows 0.600, surgeons 0.524). The mean intrarater score was 0.686 (range 0.531–0.823). A substantial agreement between rounds 1 and 2 was demonstrated in 81.8% of the raters, with a kappa score > 0.61. Stratification by level of training demonstrated similar mean intrarater coefficients (residents 0.715, fellows 0.640, surgeons 0.682). Of 570 possible questions, reviewers provided 419 correct answers (73.5%). When considering the true answer as being selected by at least one of the two main drivers of deformity, the overall accuracy increased to 86.0%. CONCLUSIONS This initial validation of a CD morphological classification system reiterates the importance of dynamic plain radiographs for the evaluation of patients with CD. The overall reliability of this CD morphological classification has been demonstrated. The overall accuracy of the classification system was not impacted by rater experience, demonstrating its simplicity.