Item Response Theory Modeling
Item Response Theory (IRT) modeling is a statistical technique that is applied after data have been collected. IRT represents the field of psychometrics -- that is, evaluation of the degree of precision and breadth of scales that are used to measure latent constructs, or underlying traits of concepts that are not directly observable and must therefore be measured indirectly. IRT consists of a class of statistical procedures that are used to model the association between an individual's responses to survey questions/items (in probabilistic terms) and an underlying latent trait that is measured by the items. IRT analysis is especially appropriate for variables such as subjective health status, treatment outcomes, and quality of life.
The results of IRT analysis can be used to determine whether scale items are appropriate for measuring a particular trait, how well items in a scale "hang together" and characterize the continuum of the underlying construct, and how strongly each of the items is connected to the underlying construct.
ARP staff have used such analysis to evaluate scales using respondent data, and this data-oriented focus of IRT largely differentiates it from other evaluation methods that are applied to questionnaire items, such as cognitive interviewing.
For a tutorial on item response theory which discusses the basics of IRT modeling and its application to health outcomes measurement, see An Introduction to Modern Measurement Theory.
Last Modified: 03 Sep 2013