Measurement Invariance of Posttraumatic Stress Disorder Symptoms Across Three Civilian Trauma Types

Abstract

The factor structure of posttraumatic stress disorder (PTSD) remains the subject of intense investigation. The DSM three-factor conceptualization of PTSD has not been empirically supported; rather, two four-factor models of PTSD (King, Leskin, King, & Weathers, 1998; Simms, Watson, & Doebbeling, 2002) have garnered the majority of support from confirmatory factor analytic (CFA) studies. Recently, interest has turned to examination of the generalizability of these well-supported models across diverse samples. Termed factorial or measurement invariance, these studies answer the question of whether PTSD maintains the same factor structure, and can be measured equivalently, across samples. However, no studies have examined PTSD‘s measurement invariance across distinct, homogeneous samples exposed to different trauma types. The current study examined the factor structure and measurement invariance of PTSD in, and across, three groups of trauma-exposed college students (N = 854) using the PTSD Checklist—Specific Version (PCL-S; Weathers, Litz, Herman, Juska, & Keane, 1993). Participants were grouped according to self-reported direct exposure to one of three distinct trauma types: motor vehicle accidents (MVA), sexual assault (SA), and sudden unexpected death of a loved one (SUD). Five models were tested using within-groups CFA, and three models showed adequate fit in each trauma group (Elhai, Biehn, Armour, Klopper, Frueh, & Palmieri, 2011; King et al., 1998; Simms et al., 2002). Multiple-group CFA showed that factor loadings were equivalent across groups for the Elhai et al. (2011) and the King et al. (1998) models, but not for the Simms et al. (2002) model. However, intercepts differed between groups in all three models. These findings suggest that PTSD symptoms as measured by the PCL-S cannot be compared across individuals who have experienced different types of traumatic events, as observed scores on the PCL-S may not correspond to levels of the latent factors equivalently across trauma types. Implications and limitations of the current results are discussed.