Cross-validation of non-memory-based embedded performance validity tests for detecting invalid performance among patients with and without neurocognitive impairment

• Published in: Journal of clinical and experimental neuropsychology Vol. 42; no. 5; pp. 459 - 472
• Main Authors: White, Daniel J., Korinek, Dale, Bernstein, Matthew T., Ovsiew, Gabriel P., Resch, Zachary J., Soble, Jason R.
• Format: Journal Article
• Language: English
• Published: England Routledge 27.05.2020; Swets & Zeitlinger bv
• Subjects: Accuracy; Adult; Cognition; Cognitive ability; Cognitive Dysfunction - diagnosis; Cross-Sectional Studies; Executive function; Executive Function - physiology; executive functioning; Female; Humans; Language; Language Tests - standards; Male; Malnutrition - diagnosis; Memory; Middle Aged; Neuropsychological Tests - standards; Outpatients; performance validity assessment; processing speed; psychometrics; Psychomotor Performance - physiology; Reproducibility of Results; Sensitivity and Specificity; Stroop Test - standards; Trail Making Test - standards; Validity; Wisconsin Card Sorting Test - standards; psychometrics; Language; executive functioning; performance validity assessment; processing speed
• ISSN: 1380-3395; 1744-411X; 1744-411X
• DOI: 10.1080/13803395.2020.1758634

Embedded performance validity tests (PVTs) allow for continuous and economical validity assessment during neuropsychological evaluations; however, similar to their freestanding counterparts, a limitation of well-validated embedded PVTs is that the majority are memory-based. This study cross-validated several previously identified non-memory-based PVTs derived from language, processing speed, and executive functioning tests within a single mixed clinical neuropsychiatric sample with and without cognitive impairment. This cross-sectional study included data from 124 clinical patients who underwent outpatient neuropsychological evaluation. Validity groups were determined by four independent criterion PVTs (failing ≤1 or ≥2), resulting in 98 valid (68% cognitively impaired) and 26 invalid performances. In total, 23 previously identified embedded PVTs derived from Verbal Fluency (VF), Trail Making Test (TMT), Stroop (SCWT), and Wisconsin Card Sorting Test (WCST) were examined. All VF, SCWT, and TMT PVTs, along with WCST Categories, significantly differed between validity groups (η p 2 =.05-.22) with areas under the curve (AUCs) of.65-.81 and 19-54% sensitivity (≥89% specificity) at optimal cut-scores. When subdivided by impairment status, all PVTs except for WCST Failures to Maintain Set were significant (AUCs =.75-94) with 33-85% sensitivity (≥90% specificity) in the cognitively unimpaired group. Among the cognitively impaired group, most VF, TMT, and SCWT PVTs remained significant, albeit with decreased accuracy (AUCs =.65-.76) and sensitivities (19-54%) at optimal cut-scores, whereas all WCST PVTs were nonsignificant. Across groups, SCWT embedded PVTs evidenced the strongest psychometric properties. VF, TMT, and SCWT embedded PVTs generally demonstrated moderate accuracy for identifying invalid neuropsychological performance. However, performance on these non-memory-based PVTs from processing speed and executive functioning tests are not immune to the effects of cognitive impairment, such that alternate cut-scores (with reduced sensitivity if adequate specificity is maintained) are indicated in cases where the clinical history is consistent with cognitive impairment. In contrast, WCST indices generally had poor accuracy.