Classification of Mandarin-speaking persons with aphasia: a discussion on validating the classification function of WAB-R

Abstract

ABSTRACT Background The classical aphasia taxonomy has been incorporated into several widely utilised aphasia assessment tools, such as the Western Aphasia Battery (WAB). Validation of the WAB’s classification typically involves comparing it with statistical clustering analysis results, using overall agreement rates as the referential index. Nevertheless, previous studies have documented diverse agreement rates. The classification function of the Mandarin version of the revised WAB (WAB-R) has not been validated. Additionally, the discussion on validating classification function is inadequate. Aim The current study adopts the argument-based validation framework to validate the classification function of the Mandarin version of the WAB-R and discuss the efficacy of using cluster analysis as evidence for the validation. Methods & Procedures A validation argument for the classification function was developed, utilising the k-means cluster analysis and within-group linkage cluster analysis as statistical procedures. WAB-R testing score data were collected from 122 Mandarin-speaking persons with acquired aphasia (non-acute). Eight clusters were expected to be generated from statistical procedures as a confirmatory test. To explore the impact of cluster numbers, solutions of generating six and ten clusters were conducted for comparative analysis. The potential reasons for mismatches were analysed. A proposed modified fluency scale (raising the cutoff score from 4 to 5) was also tested on the current dataset. Outcomes & Results Under the solution of eight clusters, the overall agreement was 62.3% with the k-means cluster analysis and 56.6% with the within-group linkage cluster analysis. Agreement rates slightly increased with the number of clusters in the within-group linkage method and remained almost unchanged in the k-means method. Anomic aphasia presented the highest agreements, while the agreements of other subtypes varied across different solutions. The fundamental differences between the clustering algorithm and the rationale of the WAB-R classification criteria were the primary reason for mismatched cases, undermining the reliability of the validation procedure. The modified fluency scale was found to significantly improve the overall agreement rates. Conclusions The classification function of the WAB-R’s Mandarin adaptation was considered acceptable under the current method. The validation argument is suggested to be modified to enhance the rationale as the statistical cluster analysis showed reliability issues when being solely employed for validation. The mismatch rate did not form direct challenges to the validity of the classification function. Additionally, modifications to the classification criteria are suggested for further exploration as the modified fluency scale has demonstrated positive results.