Do we always measure the same? On measurement invariance and response shift in rehabilitation research - part 2

• Published in: Die Rehabilitation Vol. 51; no. 6; p. 415
• Main Authors: Jelitte, M, Schuler, M
• Format: Journal Article
• Language: German
• Published: Germany 01.12.2012
• Subjects: Biomedical Research - methods; Data Interpretation, Statistical; Germany; Humans; Individuality; Outcome Assessment (Health Care) - methods; Quality of Life; Rehabilitation - methods; Rehabilitation - standards; Reproducibility of Results; Germany
• ISSN: 1439-1309
• DOI: 10.1055/s-0031-1295447

Interventions in medical rehabilitation are often evaluated using a single-group pre-post study design with health-related quality of life (hrqol) as an outcome variable. Through comparison of mean values in hrqol subscales treatment effects are calculated. In many cases conclusions about changes in hrqol are made depending on the sizes of effects. The precondition for doing this, however, is that the answers to the items in questionnaires were given within the same frames of internal standards, values and conceptualizations at the different time points. Changes in these frames can be found, however, and are discussed under the term response shift, which can happen when adjusting to chronic and progressive diseases. Existence of response shift can be proven with confirmatory factor analysis (CFA) by measuring hrqol at different time points. This approach can be assigned to the broader issue of measurement of invariance in longitudinal studies and is described in a sample of 279 patients with diabetes mellitus. Different response shift processes were detectable. If response shift takes place but is not taken into account inferences from changes in scale scores to changes in hrqol are invalid. This means that the calculation of effect size is also influenced by response shift. By the use of CFA conventionally calculated effect size can be differentiated into either effects due to response shift or 'true change' of hrqol. Measurements of invariance within one group at 2 time points can be differentiated from multiple group analysis at one time point. Investigations of measurement of invariance in longitudinal studies allow for conclusions regarding sensitivity to change of instruments examining hrqol changes. This is important for clinicians who make decisions about which scales are appropriate to detect hrqol changes. For scientific research it is relevant for further analysis of sensitivity to change of hrqol instruments.