Validity indicators within the Wisconsin Card Sorting Test: application of new and previously researched multivariate procedures in multiple traumatic brain injury samples.

The Wisconsin Card Sorting Test (WCST) is a popular neuropsychological measure of executive dysfunction that has been researched with regard to invalid performances, a subset of which in a forensic context could be associated with malingering. In the first of three studies, WCST multivariate approaches identified in prior research (Bernard, McGrath, & Houston, 1996; Suhr & Boyer, 1999), as well as newly created variables, were used to differentiate 33 chronic traumatic brain injury (TBI) patients with good effort and 27 patients with probable insufficient effort (IE). Newly created variables that were derived logically based upon hypotheses regarding strategies that might be employed by malingerers were not effective in differentiating TBI and IE groups. Application of previously researched validity indicators based upon commonly used WCST variables, individually, and within new logistic regression findings were reasonably effective in differentiating TBI and IE groups. In order to determine whether results would vary in different TBI samples, these validity indicators were examined in Study 2 with 75 moderate and severe, acute TBI rehabilitation patients whose posttraumatic amnesia had just resolved. Statistically significant differences were present between the IE group of Study 1 and the rehabilitation patient group of Study 2 on failures to maintain set, number of trials to achieve first correct category, and number of categories completed. All these measures were performed more poorly by the IE group. However, previously used multivariate approaches and the logistic regression analysis developed in Study 1 ranged widely in the degree to which Study 2's more acute rehabilitation patients were correctly classified. Specifically, the discriminant function of Bernard and colleagues correctly classified 73% and the Suhr and Boyer logistic regression correctly classified 75% of the Study 2 participants. The Study 1 logistic regression classified 97% of the Study 2 participants correctly. In Study 3, 130 mild to severe TBI patients in the VA system were studied. The Study 1 IE group performed significantly worse than the more acute and more severe VA TBI group on all 10 common WCST variables of interest. Application of the three multivariate procedures resulted in good to excellent classification rates: Suhr and Boyer logistic regression 85%, Bernard et al. discriminant function 85%, and Study 1 logistic regression 99%. The aggregate discussion of the three studies focuses on apparent differences in samples associated with varying degrees of success in identifying TBI patients. Application of these validity indicators in forensic situations should consider that some of these multivariate approaches possess possible classification limitations associated with chronicity and severity of the reported TBI. Only the Study 1 logistic regression demonstrated improved classification rates with the more acute and severe patients of Study 2 and Study 3. As with all validity indicators, use of any WCST IE criteria in isolation would not be appropriate.