Methods to detect response shift in quality of life data: a convergent validity study

Area of Science:

• Health Psychology
• Psychometrics
• Cancer Survivorship Research

Background:

• Measuring quality of life (QL) changes using pretest-posttest designs is susceptible to response shift.
• Response shift can bias results by altering participants' interpretations or evaluations of their quality of life over time.

Purpose of the Study:

• To investigate the convergent validity of three distinct methods for detecting response shift.
• To compare the thentest, anchor-recalibration, and Structural Equation Modeling (SEM) approaches in identifying response shifts in cancer patients.

Main Methods:

• The study employed a pretest-posttest design with 170 cancer patients undergoing invasive surgery.
• Participants completed the SF-36, Multidimensional Fatigue Inventory, and an anchor-recalibration task at pretest and posttest (three months after surgery).
• The thentest approach used retrospective QL judgments, anchor-recalibration assessed changes in scale-anchor definitions, and SEM analyzed factor solutions and variance-covariance matrices.

Main Results:

• The thentest and SEM approaches demonstrated strong agreement, detecting response shift in 8 out of 9 scales.
• For one scale, thentest and SEM detected response shift in opposite directions, with explanations discussed.
• The anchor-recalibration task showed limited agreement with the other methods, validating response shift absence in only 4 out of 7 scales.

Conclusions:

• The convergent findings between the thentest and SEM approaches support their validity for detecting response shift.
• These methods are reliable even when using statistically independent operationalizations of response shift.
• Recall bias did not invalidate the thentest results in this study, enhancing its utility.