Longitudinal Rasch modeling in the context of psychotherapy outcomes assessment

The present study illustrates an extension of Kamata's (2001) restricted form of the hierarchical generalized linear model that provides a multilevel longitudinal Rasch measurement model appropriate for use with polytomous responses. This extension can be used to assess average and interindividual change in the latent trait of interest, concurrently with an assessment of the invariance of item locations over time. Responses of 1,353 college students to three subscales (namely, measures of depression/anxiety, stress, and well-being) of the Outcome Questionnaire were used for demonstration purposes. Interpretation of the results is provided, and the benefits of using a multilevel item response theory model with longitudinal data are discussed. Index terms: longitudinal Rasch model, 1P HGLM, multilevel Rasch model, psychotherapy outcomes, Outcome Questionnaire, item parameter drift.