Order effects in PTSD network analysis: important implications for diagnostic conceptualization, treatment refinement, and research

Abstract

Abstract Background For decades confirmatory factor analysis (CFA) has been the preeminent method to study the underlying structure of posttraumatic stress disorder (PTSD); however, methodological limitations of CFA have led to the emergence of other analytic approaches. In particular, network analysis has become a gold standard to investigate the structure and relationships between PTSD symptoms. A key methodological limitation, however, which has significant clinical implications, is the lack of data on the potential impact of item order effects on the conclusions reached through network analyses. Methods The current study, involving a large sample (N = 5055) of active duty army soldiers following deployment to Iraq, assessed the vulnerability of network analyses and prevalence rate to item order effects. This was done by comparing symptom networks of the DSM-IV PTSD checklist items to these same items distributed in random order. Half of the participants rated their symptoms on traditionally ordered items and half the participants rated the same items, but in random order and interspersed between items from other validated scales. Differences in prevalence rate and network composition were examined. Results The prevalence rate differed between the ordered and random item samples. Network analyses using the ordered survey closely replicated the conclusions reached in the existing network analyses literature. However, in the random item survey, network composition differed considerably. Conclusion Order effects appear to have a significant impact on conclusions reached from PTSD network analysis. Prevalence rates were also impacted by order effects. These findings have important diagnostic and clinical treatment implications.