Detecting response shift in health-related quality of life measurement among patients with hypertension using structural equation modeling

Background Outcomes derived from longitudinal self-reported health-related quality of life measures can be confounded by response shift. This study was aimed to detect response shift among patients with hypertension attending a community-based disease management program. Methods 240 consecutive consulting or follow-up patients with diagnosed hypertension were recruited. The Short Form 36-item Health Survey was self-administered at 12 community health service stations at baseline and four weeks after attending the program. The 4-step structural equation modeling approach assessed response shift. Results Data from 203 (84.6%) patients were eligible for analyses (mean age 65.9 +/- 10.8 years, 46.3% female). The results showed uniform recalibration of social functioning (chi(2)(SBdiff)(1)=22.98, P < 0.001), and non-uniform recalibration of role limitations due to physical problems (chi(2)(SBdiff)(1)=8.84, P=0.003), and bodily pain (chi(2)(SBdiff)(1)=17.41, P < 0.001). The effects of response shift on social functioning were calculated as "small" (effect-size=0.35), but changed the observed changes from improvement (effect-size=0.25) to slight deterioration (effect-size=-0.10). After accounting for the response shift effect, the general physical health of participants was improved (effect-size=0.37), while deterioration (effect-size=-0.21) in the general mental health was also found. Conclusions Recalibration existed among patients with hypertension attending the disease management program. The interventions in the program might act as a catalyst that induced the response shift. We conclude that response shift should be considered in hypertension research with longitudinal health-related quality of life data.