Deriving a preference-based measure for cancer using the EORTC QLQ-C30 : a confirmatory versus exploratory approach
Costa, Daniel S.J. , King, Madeline T. , Aaronson, Neil K. , Fayers, Peter M. , Grimison, Peter S. , Janda, Monika, Pallant, Julie F. , Rowen, Donna , Velikova, Galina , Viney, Rosalie , & Young, Tracey A. (2012) Deriving a preference-based measure for cancer using the EORTC QLQ-C30 : a confirmatory versus exploratory approach. Health and Quality of Life Outcomes.
Background: To derive preference-based measures from various condition-specific descriptive health-related quality of life (HRQOL) measures. A general 2-stage method is evolved: 1) an item from each domain of the HRQOL measure is selected to form a health state classification system (HSCS); 2) a sample of health states is valued and an algorithm derived for estimating the utility of all possible health states. The aim of this analysis was to determine whether confirmatory or exploratory factor analysis (CFA, EFA) should be used to derive a cancer-specific utility measure from the EORTC QLQ-C30.
Methods: Data were collected with the QLQ-C30v3 from 356 patients receiving palliative radiotherapy for recurrent or metastatic cancer (various primary sites). The dimensional structure of the QLQ-C30 was tested with EFA and CFA, the latter based on a conceptual model (the established domain structure of the QLQ-C30: physical, role, emotional, social and cognitive functioning, plus several symptoms) and clinical considerations (views of both patients and clinicians about issues relevant to HRQOL in cancer). The dimensions determined by each method were then subjected to item response theory, including Rasch analysis.
Results: CFA results generally supported the proposed conceptual model, with residual correlations requiring only minor adjustments (namely, introduction of two cross-loadings) to improve model fit (increment χ2(2) = 77.78, p < .001). Although EFA revealed a structure similar to the CFA, some items had loadings that were difficult to interpret. Further assessment of dimensionality with Rasch analysis aligned the EFA dimensions more closely with the CFA dimensions. Three items exhibited floor effects (>75% observation at lowest score), 6 exhibited misfit to the Rasch model (fit residual > 2.5), none exhibited disordered item response thresholds, 4 exhibited DIF by gender or cancer site. Upon inspection of the remaining items, three were considered relatively less clinically important than the remaining nine.
Conclusions: CFA appears more appropriate than EFA, given the well-established structure of the QLQ-C30 and its clinical relevance. Further, the confirmatory approach produced more interpretable results than the exploratory approach. Other aspects of the general method remain largely the same. The revised method will be applied to a large number of data sets as part of the international and interdisciplinary project to develop a multi-attribute utility instrument for cancer (MAUCa).
Impact and interest:
Citation countsare sourced monthly fromand citation databases.
Citations counts from theindexing service can be viewed at the linked Google Scholar™ search.
|Keywords:||Cancer, Preference-based measure, measure, EORTC QLQ C30, confirmatory approach, exploratory|
|Subjects:||Australian and New Zealand Standard Research Classification > MEDICAL AND HEALTH SCIENCES (110000)|
|Divisions:||Current > QUT Faculties and Divisions > Faculty of Health|
Current > Institutes > Institute of Health and Biomedical Innovation
Current > Schools > School of Public Health & Social Work
|Copyright Owner:||Copyright 2012 BioMed Central|
|Deposited On:||20 Nov 2012 08:52|
|Last Modified:||20 Nov 2012 09:26|
Repository Staff Only: item control page