Keywords: generalizability theory, reliability, clinician-reported outcomes, clinical outcome assessments; measurement, dependability
Generalizability (G) theory is a statistical theory about the dependability or reliability of behavioral measurements. The strength of G theory is that multiple sources of variance (such as from persons, raters, and time) in a measurement can be estimated separately in a single analysis. In the process, G theory provides a summary coefficient reflecting the level of dependability in measurement. Originating in the behavioral and educational sciences, G theory also has merit in the reliability assessment of clinical outcome assessments in the health sciences, in particular, for clinician-rated outcomes. Yet it has been underused there. With the advent of a published guidance on clinician-rated assessments of treatment benefit by an ISPOR Task Force (Value in Health 2017), however, the application of G theory in this area is both timely and relevant. In this presentation, the fundamentals and application of G theory in the context of clinician-rated outcomes are described and illustrated.