Cutoff Elasticity in Multivariate Models of Performance Validity Assessment as a Function of the Number of Components and Aggregation Method

AbstractThere are growing concerns that increasing the number of performance validity tests (PVTs) may inflate the false positive rate. Indeed, the number of available embedded PVTs increased exponentially within the last decades. However, the standard for declaring a neurocognitive profileinvalid ( ≥ 2 PVT failures) has not been adjusted to reflect this change. Data were collected from 100 clinically referred patients with traumatic brain injury. Two distinct aggregation methods were used to combine multiple (5, 7, 9, 11 and 13) embedded PVTs into a single-number summary of performance val idity using two established free-standing PVTs as criteria. Multivariate cutoffs had to be adjusted to contain false positives: ≥ 2 failures out of nine or more dichotomized (Pass/Fail) PVTs had unexpectedly low multivariate specificity (.76-.79). However,  ≥ 4 failures resulted in high specificity (.90-.96), even out of 13 embedded PVTs. Multivariate models of embedded PVTs correctly classified between 92 and 96% of the sample at ≥ .90 specificity. Alternative aggregation methods produced similar results. Findings support the notion of th e elasticity of multivariate cutoffs: as the number of PVTs interpreted increases, more stringent cutoffs are required to deem the profileinvalid– at least until a certain level of evidence for non-credible responding accumulates (cutoff elasticity). A desirable byproduct of increasing the number of PVTs was improved sensitivity (...
Source: Psychological Injury and Law - Category: Medical Law Source Type: research