Response shift detection through then-test and structural equation modelling: Decomposing observed change and testing tacit assumptions

Assessment of change in patient-reported outcomes may be invalidated by the occurrence of response shift. Response shift refers to a change in a respondent’s internal standards that may cause changes in observed variables that are not directly related to change in the construct of interest. An established approach for detecting response shift in the area of health-related quality of life (HRQL) is to administer a retrospective pre-test (then-test). In this study, the then-test was incorporated in the structural equation modelling (SEM) approach to (1) compare the then-test approach and the SEM approach in their decomposition of observed change and (2) to test the underlying assumptions of the then-test approach. In an application to HRQL data of 170 cancer patient sundergoing invasive surgery, we found that both approaches revealed a similar pattern of decomposition, although there were some differences in the size and direction of change. With regard to the underlying assumptions of the then-test approach, results showed: (1) no evidence for recall-bias (Recall Assumption supported for all scales), (2) that internal standards of measurement were not invariant across post- and then-test measures (Consistency Assumption rejected for four out of nine scales), and (3) that internal standards were not only affected by the recalibration type of response shift (Recalibration Assumption rejected for three out of nine scales). Valid approaches for detecting response shift and the consequences assessing changes in HRQL should be further investigated.