A (five-)level playing field for mental health conditions?: exploratory analysis of EQ-5D-5L-derived utility values

E. M. Camacho, G. Shields, K. Lovell, P. A. Coventry, A. P. Morrison, L. M. Davies

Economic evaluations of mental health interventions often measure health benefit in terms of utility values derived from the EQ-5D. For the five-level version of the EQ-5D, there are two methods of estimating utility [crosswalk and stated preference (5L-SP)]. This paper explores potential impacts for researchers and decision-makers when comparing utility values derived from either method in the specific context of mental health. Baseline EQ-5D-5L data from three large randomised controlled trials of interventions for mental health conditions were analysed. Utility values were generated using each method. Mean utility values were compared using a series of t tests on pooled data and subgroups. Scenario analyses explored potential impacts on cost-effectiveness decisions. EQ-5D data were available for 1399 participants. The mean utility value for each trial was approximately 0.08 higher when estimated using the 5L-SP approach compared to crosswalk (p < 0.0001). The difference was greatest among people reporting extreme anxiety/depression (mean utility 5L-SP 0.309, crosswalk 0.084; difference = 0.225; p < 0.0001). Identical improvements in health status were associated with higher costs to gain one QALY with the 5L-SP approach; this is more pronounced when improvements are across all domains compared to improvements on the anxiety/depression domain only. The two approaches produce significantly different utility values in people with mental health conditions. Resulting differences in cost per QALY estimates suggest that thresholds of cost-effectiveness may also need to be reviewed. Researchers and decision-makers should exercise caution when comparing or synthesising data from trials of mental health interventions using different utility estimation approaches.