151. Testing the measurement invariance of the EORTC QLQ-C30 across primary cancer sites using multi-group confirmatory factor analysis
- Author
-
Madeleine King, Galina Velikova, Daniel S.J. Costa, Peter Fayers, Julie F Pallant, Neil K. Aaronson, and Klinische Psychologie (Psychologie, FMG)
- Subjects
Adult ,Male ,Multivariate analysis ,Breast Neoplasms ,Breast cancer ,Sickness Impact Profile ,Surveys and Questionnaires ,Post-hoc analysis ,Statistics ,medicine ,Humans ,Measurement invariance ,Equivalence (measure theory) ,Aged ,Factor analysis ,Models, Statistical ,Confounding ,Public Health, Environmental and Occupational Health ,Reproducibility of Results ,Middle Aged ,medicine.disease ,humanities ,Confirmatory factor analysis ,Multivariate Analysis ,Quality of Life ,Female ,Factor Analysis, Statistical ,Psychology ,Social psychology - Abstract
Purpose: The EORTC Quality of Life Questionnaire is a widely used cancer-specific quality of life instrument comprising a core set of 30 items (QLQ-C30) supplemented by cancer site-specific modules. The purpose of this paper was to examine the extent to which the conventional multi-item domain structure of the QLQ-C30 holds across patients with seven different primary cancer sites.Methods: Multi-group confirmatory factor analysis was used to test whether a measurement model of the QLQ-C30 was invariant across cancer sites. Configural (same patterns of factor loadings), metric (equivalence of factor loadings) and scalar (equivalence of thresholds) invariance amongst the cancer site groups were assessed (N = 1,906) by comparing the fit of a model with these parameters freely estimated to a model where estimates were constrained to be equal for the corresponding items in each group.Results: All groups exhibited good model fit except for the prostate group, which was excluded. Only 1 of 576 parameters was found to differ between primary sites: specifically, the first threshold of Item 1 in the breast cancer group exhibited non-invariance. In a post hoc analysis, several instances of non-invariance by treatment status (baseline, on-treatment, off-treatment) were observed.Conclusions: Given only one instance of non-invariance between cancer sites, there is a reason to be confident in the validity of conclusions drawn when comparing QLQ-C30 domain scores between different sites and when interpreting the scores of heterogeneous samples, although future research should assess the potential impact of confounding variables such as treatment and gender.
- Published
- 2014
- Full Text
- View/download PDF