The Oort structural equation modeling approach detected a response shift after a COPD self-management program not detected by the Schmitt technique

详细信息    查看全文

• 作者：Sara Ahmed ; Jean Bourbeau ; Franç ; ois Maltais ; Asmaa Mansour
• 刊名：Journal of Clinical Epidemiology
• 出版年：2009
• 出版时间：November 2009
• 年：2009
• 卷：62
• 期：11
• 页码：1165-1172
• 全文大小：166 K

文摘

Objective

If individuals experience a response shift, scores on measures before and after a self-management intervention may not be comparable. This study evaluated whether persons with chronic obstructive pulmonary disease (COPD) experience a response shift after participating in a self-management program. The second objective was to compare the Oort and Schmitt structural equation modeling (SEM) approaches.

Methods

Secondary analyses from a randomized controlled trial comparing a home- and hospital-based pulmonary rehabilitation program were used to assess response shift on a physical and mental health–measurement model measured using the Chronic Respiratory Questionnaire (CRQ) and the St. George's Respiratory Questionnaire (SGRQ) over a 1-year period.

Results

The Oort approach showed significant changes between the no response shift model and models removing invariance constraints for the residual of the CRQ dyspnea (χ²_SBdiff = 7, df = 1) (uniform recalibration) and intercepts of the SGRQ activity (χ²_SBdiff = 14, df = 1) and impact (χ²_SBdiff = 10, df = 1) subscales (nonuniform recalibration). Change in factor means showed changes in the physical health factor, which was slightly lower in unadjusted (0.32) as compared with the response shift–adjusted model (0.40). The Schmitt procedure was not supportive of any response shift effect and showed a marginal change in random error over time.

Conclusions

This study showed that COPD patients experienced a response shift after participating in a self-management program, which resulted in an underestimation of change in physical health. These results suggest that the Oort procedure is more sensitive in detecting a response shift, and that a measurement of response shift is needed before results can be interpreted. Future comparisons with other methods and a control group are needed.