Emergence of item response modeling in instrument development and data analysis

In summary, readers are encouraged to read carefully the IRT articles in this special issue. They provide theoretical as well as practical arguments in favor of using IRT models in the health outcomes measurement field. At the same time, readers should be aware that the IRT field is complex and software is limited and, in my judgment, not very user friendly (although some packages are better than others). In addition, sample sizes will often need to be larger than in classical measurement, at least with the more general IRT models, and applications are rarely straightforward. Considerable practical experience is needed to ensure successful applications of IRT in the development and validation of instruments for health outcomes measurement.