On the psychometric validity of the domains of the PDSQ: An illustration of the bi-factor item response theory model

Abstract

Competing item response theory (IRT) models were used to test the factor structure of the psychiatric diagnostic screening questionnaire (PDSQ; Zimmerman M, Mattia JI. A self-report scale to help make psychiatric diagnoses: the psychiatric diagnostic screening questionnaire. Archives of General Psychiatry 2001;58:787–94), a self-report psychiatric measure comprised of 139 items sampled from 15 symptom domains (e.g., Psychosis, Mania). Tested IRT models included: (a) a unidimensional model, (b) a simple structure model, (c) a bi-factor model, and (d) models that included 6, 10, and 15 sub-domain alternative conceptualizations of the scale. Based on the responses of 3791 individuals with major depressive disorder, the bi-factor model was found to provide a theoretically and statistically plausible description of the PDSQ factor structure. Primary dimension loadings were low to moderate; group factor loadings were moderate to high. Results support the validity of the PDSQ in identifying distinct categories of illness as defined by the diagnostic and statistical manual diagnostic groups, since preserving the 15 symptom categories (domains) provided a more accurate characterization of the observed data by the IRT models. The bi-factor model is useful in evaluating the multidimensional structure of mental health data. The specification of alternative IRT models is demonstrated as a noteworthy benefit over classical test theory for psychiatric measurement.