Online Program

Item Response Theory Approaches for Research Synthesis and Harmonization

*Robert D Gibbons, The University of Chicago 

Keywords: individual participant data, harmonization, meta-analysis

A common problem in research synthesis is the desire to analyze person-level longitudinal outcomes that are measured using different instruments (e.g. rating scales) in different studies. The typical approach is to resort to meta-analysis in which an overall effect size is synthesized across studies, or to use predicted total scale scores based on the calibration of one instrument versus another. An alternative approach, explored in this presentation, is to use item response theory to equate the two or more tests at the item level and then generate outcome scores based on the available item responses that were obtained in the specific study. We explore use of this strategy as it relates to problems in research synthesis and harmonization and apply it to a relevant example dataset.