Study design: Multicenter observational survey study.
Objectives: To quantify and compare the inter- and intraobserver reliability of Allen-Fergusson (A-F), Harris, Argenson, and AOSpine (AOS) classifications for cervical spine injuries, in a multicentric survey of neurosurgeons with different levels of experience.
Methods: We used data of 64 consecutive patients. Totally, 37 surgeons (from 7 centers), were included in the study. The initial assessment was returned by 36 raters. The second assessment performed after 1.5 months included 24 raters.
Results: We received 15 111 answers for 3840 evaluations. Raters reached a fair general agreement of the A-F scale, while the experienced group achieved κ = 0.39. While all groups showed moderate interrater reliability for primary assessment of Harris scale (κ = 0.44), the κ value for experts decreased from 0.58 to 0.49. The Argenson scale demonstrated moderate and substantial agreement among all raters (κ = 0.47 and κ = 0.55, respectively). The AOS scheme primary assessment general kappa value for all types of injuries and across all raters was 0.49, reaching substantial agreement among experts (κ = 0.62) with moderate agreement across beginner and intermediate groups (κ = 0.48 and κ = 0.44, respectively). The second assessment general agreement kappa value reached 0.56.
Conclusions: We found the highest values of interobserver agreement and reproducibility among surgeons with different levels of experience with Argenson and AOSpine classifications. The AOSpine scale additionally incorporated more detailed description of compression injuries and facet-joint fractures. Agreement levels reached for Allen-Fergusson and Harris scales were fair and moderate, respectively, indicating difficulty of their application in clinical practice, especially by junior specialists.
Keywords: AOSpine classification; Allen-Fergusson classification; Argenson classification; Fleiss’ kappa; Harris classification; cervical spine trauma; interobserver agreement; intraobserver agreement.