Evaluating constructs represented by symptom validity tests in forensic neuropsychological assessment of traumatic...

This study uses a new method to summarize diagnostic validity information to explore which constructs are captured by malingering tests. The Test Validation Summary applies mixed-groups validation to investigate the meaning of test constructs and to estimate test classification characteristics when test validation groups are not "pure" criterion groups (ie, "compliant" vs "malingering"), but members have variable probability of malingering. The method permits the use of tests with relatively low validity to validate tests of greater validity. In our initial analysis, we argue that the Rey 15-Item Memory Test is best construed as an "intention test" (capturing the intention of testtakers when taking a test) as opposed to an "effort test." Using the Test Validation Summary and mixed-groups validation, we demonstrate that as an indicator of "intention to feign cognitive impairment," the Rey 15-Item Memory Test has estimated false-positive rate (FPR) = 0.02 and true-positive rate (TPR) = 0.57. We then explore the meaning of failure on the Word Memory Test (WMT), which uses a dichotomous classification of performance as valid or invalid. Although the WMT is commonly referred to as an "effort test," we argue that it likely captures both "intention" and "effort" but collapses this information into a single dichotomous classification of symptom validity. We demonstrate that, as a result of this dichotomous classification process, the WMT likely has a problematic FPR. In our analysis of previously published WMT data, the WMT FPR is estimated at 0.12 when there is no predisposition to perform poorly but rises dramatically and unrealistically as the predisposition to perform poorly increases. We compare these findings to those of the Validity Indicator Profile (VIP), which captures both intent and effort to classify 4 different sorts of response styles in cognitive testing. In our analyses, the VIP demonstrates that FPR = 0 and TPR = 0.86 when the construct being measured is "intent to perform poorly," and reveals that FPR = 0.06 and TPR = 0.63 when the construct being measured is "inconsistent responding" or "poor effort." We were able to demonstrate for the VIP the same "oversensitivity" shown by the WMT when the VIP was interpreted only as a dichotomous classification test. These results indicate that researchers who attempt to generate classification characteristics for malingering tests must carefully consider what constructs are being captured by the test.