Measuring change in quality of life: bias in prospective and retrospective evaluation.

OBJECTIVES: Treatment effects on health-related quality of life (QOL) often differ depending on whether they are measured prospectively (before and after treatment) or retrospectively (after treatment only). These two approaches can be subject to different sorts of bias: Prospective evaluations may be biased by scale recalibration (a changed understanding of the response scale), and retrospective evaluations may be biased by recall bias (a wrong assessment of former QOL).
METHODS: On the basis of an analysis of the literature, we present an overview on possible biases in prospective and retrospective measurement of QOL and how these biases are named and defined in the literature.
RESULTS: The definitions of different biases are inconsistent. Many authors do not clearly distinguish measurement bias from true change. Furthermore, some consider only scale recalibration or only recall bias.
CONCLUSIONS: Much of the current discussion on bias in prospective and retrospective QOL measurement suffers from unclear definitions, especially of "response shift" and "recall bias," or from neglecting one of the possible biases. We suggest more elaborate definitions for different types of bias and recommend taking both kinds of bias into consideration when measuring change in QOL. The relevance of the different biases depends on the type of study, and so either prospective or retrospective assessment may be more appropriate.