Construct validity of performance validity tests across multiple mixed-clinical samples.

OBJECTIVE: Few studies have systematically assessed construct validity across a diverse set of performance validity tests (PVTs). Clarifying whether PVTs measure a common construct or distinct aspects of performance validity has important clinical implications for neuropsychological data interpretation and PVT selection. METHODS: This study evaluated the latent structure of PVTs using retrospective data from multiple mixed-clinical samples (Sample 1: N = 115, Sample 2: N = 297; Sample 3: N = 160), with diverse mental health and neurocognitive diagnoses. Each sample was administered a variety of embedded and freestanding PVTs, and the latent structure was examined using exploratory factor analyses (Samples 1-2) and confirmatory factor analysis (CFA; Sample 3). RESULTS: Initial analyses supported a unidimensional structure (Sample 1), whereas Samples 2-3 provided converging evidence for a two-factor solution characterized by moderately to strongly correlated Memory and Non-Memory factors (r = .65 to .71) and excellent model fit in Sample 3 (robust comparative fit index = .99, robust root mean square error of approximation = .045, standardized root mean square residual = .038). Memory-based PVTs demonstrated stronger and more consistent loadings than non-memory PVTs across samples. CONCLUSIONS: Findings were consistent with a hierarchical organization with a general PVT construct at the highest strata and closely related Memory and Non-Memory factors nested beneath. Alternative design-based explanations are also considered. Clinically, results support the continued use of multiple memory and non-memory-based PVTs throughout a neuropsychological assessment. Future replication is needed using a prospective design with larger, more demographically and clinically diverse samples.