A-182 Development of a PTSD-Specific Embedded Validity Measure in the Personality Assessment Inventory

Abstract

Abstract Objective/Background The present study investigated the feasibility of developing a PTSD-specific malingering scale embedded within the Personality Assessment Inventory (PAI). Methods Participants consisted of 489 individuals [mean age = 20.0 (SD = 3.29); 71% female; 73.0% Caucasian, 17.6% African American, 3.5%, 9.4% Other] who completed PAI. 274 individuals were in the control group and instructed to answer normally while 215 were instructed to simulate PTSD with intent to successfully feign PTSD in the context of a legal case. The following were identified as candidate PAI subscales for logistic regression to identify feigned PTSD: ARD-T, BOR-A, SCZ-T, ANX-P, and DEP-P. Results Logistic regression analysis was performed with known group as the dependent variable and PAI subtest normed T-scores as predictor variables. The final full model of five predictor variables significantly predicted group status (χ2 = 384.9, df = 5, N = 489, p < 0.001). The model accounted for 54.5% to 73.0% of the variance (Cox and Snell R2 = 0.545; Nagelkerke R2 = 0.730) in performance classification with overall 89.4% of individuals correctly predicted to their known group. At a cut value of 0.65 the specificity was 94.9%, and the sensitivity was 82.3%. Within the model, the following variables were incrementally predictive of group membership: ARD-T (B = 0.05, p < 0.001), BOR-A (B = 0.04, p < 0.05), and ANX-P (B = 0.04, p < 0.05). Conclusions Preliminary evidence for use of an embedded PAI malingering scale within the PAI was found. Further replication is needed in clinical populations prior to use in practice.