Estimating a preference-based index from the Clinical Outcomes in Routine Evaluation-Outcome Measure (CORE-OM): valuation of CORE-6D

Abstract

Background: The Clinical Outcomes in Routine Evaluation-Outcome Measure (CORE-OM) is used to evaluate the effectiveness of psychological therapies in people with common mental disorders. The objective of this study was to estimate a preference-based index for this population using CORE-6D, a health state classification system derived from the CORE-OM consisting of a 5-item emotional component and a physical item, and to demonstrate a novel method for generating states that are not orthogonal.

Methods: Rasch analysis was used to identify 11 emotional health states from CORE-6D that were frequently observed in the study population and are, thus, plausible (in contrast, conventional statistical design might generate implausible states). Combined with the 3 response levels of the physical item of CORE-6D, they generate 33 plausible health states, 18 of which were selected for valuation. A valuation survey of 220 members of the public in South Yorkshire, United Kingdom, was undertaken using the time tradeoff (TTO) method. Regression analysis was subsequently used to predict values for all possible states described by CORE-6D.

Results: A number of multivariate regression models were built to predict values for the 33 health states of CORE-6D, using the Rasch logit value of the emotional state and the response level of the physical item as independent variables. A cubic model with high predictive value (adjusted R(2) = 0.990) was selected to predict TTO values for all 729 CORE-6D health states.

Conclusion: The CORE-6D preference-based index will enable the assessment of cost-effectiveness of interventions for people with common mental disorders using existing and prospective CORE-OM data sets. The new method for generating states may be useful for other instruments with highly correlated dimensions.

Figure 1

Rasch item threshold map of the emotional component of CORE-6D, from Mavranezouli et al., Quality of Life Research 2011; 20(3): 321-33, reprinted with kind permission from Springer Science + Business Media. 0 = never; 1 = only occasionally or sometimes; 2 = often, most or all the time; note that the fourth item is positively worded and therefore response levels are reversed.

Figure 2

Mean observed (from the valuation survey) and modeled (based on regression model 7) time tradeoff values by Rasch rescaled logit value. Modelled time trade-off (TTO) values are predicted using the Rasch rescaled logit value of the emotional health state and the response level of the physical item ‘I am troubled by aches, pains, physical problems’ (level 0 = never; level 1 = only occasionally or sometimes; level 2 = often, most or all the time).