Selecting Bolt-on Dimensions for the EQ-5D: Testing the Impact of Hearing, Sleep, Cognition, Energy, and Relationships on Preferences Using Pairwise Choices

Background Generic preference-based measures (GPBMs) such as the EQ-5D are valid across many conditions, but in some cases, "bolting on" additional dimensions may improve validity. The selection of "bolt-ons" has been based on the psychometric impact of individual dimensions, but preferences provide another important way to select them. This study aims to test the potential of using pairwise choices to inform the selection of bolt-ons for the EQ-5D-5L. Methods General population preferences were collected using an online survey of 1040 UK residents. Three EQ-5D-5L health state pairs were selected based on pairs that had a 50:50 split in respondent preferences from a previous pairwise survey. Participants were presented with pairwise choices of EQ-5D-5L health states without and with bolt-ons of hearing, sleep, cognition, energy, and relationships, each added individually. Logistic models were used to assess the impact of bolt-ons, as well as bolt-ons at different severity levels, on the log odds of responders choosing between health states. Results Preferences varied according to the bolt-ons and their severity level (only levels 1, 3, and 5 were used). Additions of bolt-ons at level 1 generally resulted in nonstatistically significant differences while additions of bolt-ons at level 3 and level 5 produced a negative and statistically significant impact on preferences for the health state with the bolt-on. At level 5, hearing had the largest impact, followed by cognition, relationships, energy, and sleep. At level 3, cognition produced the largest impact, followed by hearing and sleep with similar impacts, energy, and relationships. This ordering offers information for bolt-on selection, with hearing and cognition appearing as the most important. The weight placed on the different health problems is not constant across severity levels between bolt-ons. Conclusions Pairwise choices provide a cost-effective approach of generating information on preferences to support bolt-on selection.