The simultaneous valuation of states from multiple instruments using ranking and VAS data: methods and preliminary results

Background: Previous methods of empirical mapping involve using regressions on patient or general population self-report data from datasets involving 2 or more instruments. This approach relies on overlap in the descriptive systems of the measures, but key dimensions may not be present in both measures. Furthermore, this assumes it is appropriate to use different instruments on the same population, which may not be the case for all patient groups. The aim of the study described here is to develop a new method of mapping using general population preferences for hypothetical health states defined by the descriptive systems of different measures. This paper presents a description of the methods used in the study and reports on the results of the valuation study including details about the respondents, feasibility and quality (e.g. response rate, completion and consistency) and descriptive results on VAS and ranking data. The use of these results to estimate mapping functions between instruments will be presented in a companion paper.

Methods: The study used interviewer administered versions of ranking and VAS techniques to value 13 health states defined by each of 6 instruments: EQ-5D (generic), SF-6D (generic), HUI2 (generic for children), AQL-5D (asthma specific), OPUS (social care specific), ICECAP (capabilities). Each interview involved 3 ranking and visual analogue scale (VAS) tasks with states from 3 different instruments where each task involves the simultaneous valuation of multiple instruments. The study includes 13 health and well-being states for each instrument (16 for EQ-5D) that reflect a range of health state values according to the published health state values for each instrument and each health state is valued approximately 75-100 times.

Results: The sample consists of 499 members of the UK general population with a reasonable spread of background characteristics (response rate=55%). The study achieved a completion rate of 99% for all states included in the rank and rating tasks and 94.8% of respondents have complete VAS responses and 97.2% have complete rank responses. Interviewers reported that it is doubtful for 4.1% of respondents that they understood the tasks, and 29.3% of respondents stated that they found the tasks difficult. The results suggest important differences in the range of mean VAS and mean rank values per state across instruments; for example, mean VAS values for the worst state vary across instruments from 0.075 to 0.324. Respondents are able to change the ordering of states between the rank and VAS tasks and 12.0% of respondents have one or more differences in their rank and VAS orderings for every task.

Conclusions: This study has demonstrated the feasibility of simultaneously valuing health states from different preference-based instruments. The preliminary analysis of the results presented here provides the basis for a new method of mapping between measures based on general population preferences.