Reliability of the classification for assessing hand behind the back following reverse shoulder arthroplasty

Hand behind the back (HBB) after reverse shoulder arthroplasty (rTSA) is important for activities of daily living (ADLs), but the commonly used vertebral-level measurements are inconsistent and only moderately reliable. Collin et al described 3 functional HBB motion patterns that capture how patients reach posteriorly (fluently vs., assisted) rather than how high they reach. We adapted this model into a bedside tool, the "HBB Classification," and compared its interobserver reliability with the traditional vertebral-level method.

In this prospective, single-center, observational cohort study, 116 shoulders that had undergone rTSA were assessed during routine post-operative follow-up visits. HBB function was evaluated by 2 blinded observers (a fellowship-trained shoulder surgeon and a physiotherapist) using 2 classification systems during the same clinical session: (1) the traditional vertebral-level method and (2) the HBB classification (Type 1: unable to achieve HBB; Type 2: HBB achieved but not fluent or assisted; Type 3: fluent, unassisted). Demographic data and surgical approach were recorded. Interobserver reliability was assessed using percent agreement and Cohen's kappa coefficient, and kappa values were compared using Fleiss' Z test, with statistical significance set at P < .05. Post-operative functional outcomes (Constant score, Simple Shoulder Test, Activities of Daily Living External and Internal Rotations, and Subjective Shoulder Value) were also compared across HBB categories.

Mean age at surgery was 70.7 ± 7.5 years (range, 52-91 years). The cohort included 116 shoulders from 110 patients; 65 patients (59.1%) were women, and 86 shoulders (74.1%) were right-sided. Mean follow-up was 15.8 ± 14.7 months (range, 6-83 months). Indications were cuff-tear arthropathy 36 (31.0%), primary osteoarthritis 31 (26.7%), massive rotator cuff tear 33 (28.4%), sequelae of fracture 7 (6.0%), and other/revision indications 9 (7.9%). Interobserver agreement for the vertebral-level method was 66.4%, with κ = 0.599 (95% CI 0.49-0.69; moderate). Agreement for the HBB classification was 96.6%, with κ = 0.945 (95% CI 0.90-0.98; almost perfect). The difference in κ was Δκ = 0.346, and was statistically significant (Z = 6.09, P < .000001). HBB categories were also significantly associated with post-operative functional outcome measures (Simple Shoulder Test, Activities of Daily Living External and Internal Rotations, Constant score, and Subjective Shoulder Value).

The HBB classification showed almost-perfect interobserver reliability and significantly outperformed the traditional vertebral-level method. As a clinical adaptation of the Southard three-pattern framework, it provides a practical, functionally meaningful, and highly reproducible way to report HBB after rTSA while its significant association with post-operative functional outcome measures further supports its clinical relevance.