The Equivalence of Traditional and Diagnostic Classification Model Approaches in Analyzing Empirical and Simulated Posttraumatic Stress Disorder Data

The purpose of this study was to evaluate the performance of a log-linear cognitive diagnosis model (LCDM) as compared to three traditional scoring methods for the diagnosis of posttraumatic stress disorder (PTSD) as measured by the Posttraumatic Stress Disorder Checklist (PCL) civilian and military versions. Both civilian (empirical) and military (simulated) data were used in this study. Changes in the measurement and structure of PTSD have occurred over the years, especially as it related to military Veterans. Just recently in 2013, the diagnostic criteria for PTSD was changed to a four-dimensional diagnosis, although this study used the pre-2013 three-dimensional diagnosis due to the time period and instruments used to collect the empirical sample and the original samples used as parameters for the simulated data. Model parameters, respondent estimates, respondent classifications, and rates of agreement between the diagnostic methods were explored for both empirical and simulated data and further for sample size differences with the simulated data. The LCDM structural model was found to be comparable to the CFA model and the average item intercept -.539, showing that roughly 37% of respondents who were not positive for any PTSD clusters endorsed the item as being a problem they experienced. The average main effect parameter was 2.318, the size of which is relative to the size of the intercept where the lower intercepts and higher main effects showed more discrimination between those respondents who were ill or not ill with PTSD. Since posttraumatic stress disorder is both a compensatory and non-compensatory diagnosis the LCDM was evaluated on how well it was able to reflect these dimensions in the respondent estimates and classifications. Overall, the LCDM performed poorly compared to the traditional scoring, with the LCDM performing better at the symptom cluster level than at the diagnostic level. The potential utility of this feedback to a practicing clinician was found to be not yet suitable to clinical practice, but this study shows an opportunity to refine the PCL instrument to perform better in the LCDM framework, or to develop an instrument that is more discriminating and may be useful in differential diagnosis