Special Issue on Item Response Theory in Medical Studies

Statistical Methods in Medical Research presents a Special Issue on IRT in Medical Studies where new developments in this research area are highlighted.

(Link To Special Issue (Open Access))


Item Response Theory (IRT) comprises a class of latent variable models and associated statistical procedures that connects observed item responses to an underlying construct. IRT models are used primarily to measure a continuous latent variable from observed categorical item observations. The latent variable represents an underlying construct that cannot be measured directly, but it is estimated from the observed item responses. IRT provides a way to construct a common measurement scale on which the latent variable scores can be represented. Applications of IRT can be found in educational, social and health sciences, and include the measurement of quality-of-life and the assessment of health status.
New technologies have stimulated further the usefulness of IRT. Large-scale assessment programs have been developed using digital technology (e.g., tablets, smartphones) to collect response data. This development created opportunities to implement computer adaptive testing and to improve assessments in terms of item and test design, data analysis, and reporting of test results. For instance, the Patient-Reported Outcome Measurement Information System (PROMIS) comprehends a collection of freely available evaluation assessments for physical, mental and social domains. The PROMIS instruments provide a way to obtain standardized outcome measures and to monitor patient outcomes using IRT.


They are operationalized on a tablet computer, which is simple and convenient in a healthcare setting. With the ever-increasing availability of computer devices (smart devices), computer-assisted assessments are becoming the standard. Together with the increase in data storage opportunities, computer-assisted test administration becomes applicable for clinical trials and longitudinal observational studies. These developments also created new computational, mathematical and statistical challenges. Due to the innovative types of data collection, more and more data are collected with complex dependence structures. To make reliable and accurate decisions from observed data, more general IRT models are needed to account for dependencies that go beyond the well-known data correlations caused by the clustering of item responses within persons and items.


Leave a Reply