Application of Rasch Analysis to Examine Psychometric Aspects of the Activities-Specific Balance Confidence Scale When Used in a New Cultural Context

Arnadottir SA, Lundin-Olsson L, Gunnarsdottir ED, Fisher AG. Application of Rasch analysis to examine psychometric aspects of the Activities-Specific Balance Confidence Scale when used in a new cultural context. Arch Phys Med Rehabil 2010;91:156-63. Objective: To investigate by using Rasch analysis the psychometric properties of the Activities-Specific Balance Confidence (ABC) Scale when applied in a new Icelandic context. Design: Cross-sectional, population-based, random selection from the Icelandic National Registry. Setting: Community-based. Participants: Icelanders (N=183), 65 to 88 years old, and 48% women. Interventions: Not applicable. Main Outcome Measure: ABC, an instrument used to evaluate how confident older people are in maintaining balance and remaining steady when moving through the environment. An Icelandic translation of the ABC (ABC-ICE) scale was evaluated by implementing Rasch rating scale analysis to transform ordinal ABC-ICE scores into interval measures and evaluating aspects of validity and reliability of the scale. Results: Participants were not able to differentiate reliably between the 11 rating scale categories of the ABC-ICE. Additionally, 3 items failed to show acceptable goodness of fit to the ABC-ICE rating scale model. By collapsing categories and creating a new 5-category scale, only I item misfit. Removing that item resulted in a modified version of ABC-ICE with 5 categories and 15 items. Both item goodness-of-fit statistics and principal components analysis supported unidimensionality of the modified ABC-ICE. The ABC-ICE measures reliably separated the sample into at least 4 statistically distinct strata of balance confidence. Finally, the hierarchical order of item difficulties was consistent with theoretic expectations, and the items were reasonably well targeted to the balance confidence of the persons tested. Conclusions: Rasch analysis indicated a need to modify the ABC-ICE to improve its psychometric properties. Further studies are needed to determine if similar analyses of other versions of the ABC, including the original one, will yield similar results.