Adding meaning to measurement: Initial rasch analysis of the ASHA FACS social communication subtest

Background: The American Speech-Language-Hearing Association Functional Assessment of Communication Skills (ASHA FACS) has established validity and reliability for individuals with aphasia. However, subtest scores are summed and averaged across subtests, making it difficult to describe individuals' communication behaviours. Item-response theory provides a methodology to assess the measurement properties of established assessments by transforming ordinal data into interval data, calibrated for both person ability measures and item difficulty on the same linear scale. The result is that behaviours can be ordered and described in ways that are more informative and useful than reporting raw scores. Aims: The study aimed to use Rasch analysis to determine the psychometric properties of the ASHA FACS Social Communication Subtest (SCS) and to demonstrate how the Key form (a visual representation of the measure that results from the Rasch analysis) could add meaning to scoring and tracking progress. Methods & Procedures: A total of 130 caregivers (76% female, 24% male) rated the performance of individuals with aphasia on the ASHA FACS SCS (21 items rated on a 7-unit scale). Participants' ages ranged from 31 to 88 years (M=60.5, SD=12.4). Rasch analysis was done by WINSTEP computer software. Results: The ASHA FACS SCS demonstrated good measurement characteristics with two exceptions. First, three of seven units of the rating scale were used with such low probability that they were not useful to the measure. The data were re-analysed using the four-unit scale. The new rating scale covered the full range of sample abilities (2.94 statistically distinct levels). Second, the principal components analysis (an analysis of unidimensionality) found negative factor-item loading of easier items (understanding) and positive factor-item loading of harder items (conversing), indicating that the scale contained two constructs rather than one. Fit statistics showed all 21 items within the established criteria (.6 < Mean Square < 1.4 ZStd < 2.0). The person separation reliability (analogous to Cronbach's alpha) was high at .90. The Key form linked the measure of item difficulty (logits) and rating scale with the item hierarchy. Conclusions: Through Rasch analysis of the ASHA FACS SCS we demonstrated that: caregivers were reliable respondents (adding another use for the scale); the rating scale could be simplified; and the Key form could be used to demonstrate treatment progress. Future investigation could include: (1) analysis of the measurement properties of all subtests based on a large sample of caregivers; (2) item development for two constructs of social communication (e.g., understanding and conversing); and (3) test-retest reliability of respondents.