Minimal evidence of response shift in the absence of a catalyst

Individuals with chronic conditions experience fluctuations in health status and thus may experience response shift. We sought to test the hypothesis that response shift effects would be non-significant among individuals with chronic disease who experienced relatively small changes in their health status over a 1-year period.

This secondary analysis utilized longitudinal cohort data on a community-based sample (n = 776) representing four chronic diseases (arthritis, heart failure, diabetes, or chronic obstructive pulmonary disease). Information on health-care utilization was obtained from the provincial health insurance database. Participants completed the SF-36 twice annually. Parameter invariance over 1 year in a second-order SF-36 factor structure was evaluated by adapting Oort's approach by fitting a second-order measurement structure with first-order factors for the SF-36 subscales and second-order factors for physical and mental health status while accommodating ordinal data.

Over 80 % of participants had no hospitalizations or emergency room visits over follow-up. The model had an acceptable fit when all measurement model parameters were constrained at both time points (RMSEA = .035, CFI = .97). There was no substantial difference in fit when measurement model parameters (item thresholds, first-order factor intercepts, and factor loadings) were allowed to vary over time.

Among chronically ill individuals with stable health, substantial response shift effects were not detected. These results support the theoretical proposition that response shift is not expected to occur in patients with relatively stable conditions.