Level of Agreement Between Human-Rated and Instrumented Balance Error Scoring System Scores

Clinicians have used the Balance Error Scoring System (BESS) to quantify postural control for concussion management. However, the reliability of the human rated BESS has varied prompting the development of instrumented BESSs. A cross-sectional design was used to determine the level of agreement (LOA) between human rated and instrumented BESS scores. Sixty participants completed the BESS on video. An instrumented mat was used to quantify BESS errors while a live human rater simultaneously scored the BESS. A second human rated BESS performance via video. Bland-Altman LOA analyses evaluated agreement between scoring methods (Mat-Human, Mat-Video, Video-Live) for each stance. Mean biases between scores, for each stance, with 95% confidence intervals (95%CIs) were calculated. Agreement between scoring methods was not assessed for the Firm-Double-Limb stance because very few errors were recorded. Agreement between both human raters and the mat was poor based on mean bias estimates > ± 1 and/or wide 95%CIs for all stances including BESS-Total. Agreement between the human raters was better, having displayed consistently smaller mean bias estimates and tighter 95%CIs for all stances and BESS Total. As a result, human rated and instrumented BESS scores may not be comparable. One method should be used to measure BESS errors for consistency.