[Analyzing the Informational Content of Observational and Rating Data]

In rehabilitation, decisions are often based on ratings or judgements made by rehabilitation patients, therapists or peer-reviewers. But, depending on aspects of the assessment setting, ratings of an issue (e. g., quality of communication, quality of health care) may vary systematically or due to stochastic error components. Rating scores may be affected by the properties of the rater, raters' perspective (e. g., self- vs. external ratings), rating situation or measurement point in time, respectively. Hence, rating scores should be considered as composite scores comprising several systematic information aspects. Generalizability theory provides an analytical framework allowing to decompose underlying information facets and to determine their influence on the obtained rating data. First, a differentiated model can be identified ensuring a more valid data interpretation (generalizability aspect). Second, well-founded recommendations enhancing reliability and validity of the assessment may be derived (decision aspect). In this article we demonstrate how to apply generalizability theory to enhance the clarity of the informational content of ratings and to foster rating quality in typical rehabilitation settings.