Detailed histories of the EuroQol Group and its development of the EQ-5D are available elsewhere [1–3]. In this paper the focus is on the following questions: Why was the EQ-5D initially developed? How did its use and application evolve? How will the EuroQol Group proceed?
The Provenance of the EQ-5D
The EuroQol Group started its journey in 1987: 14 people met to exchange ideas about how to approach the development of a health status measurement instrument. One of the motivations for doing so was to assist healthcare decision-makers to make resource allocation decisions informed by evidence on the cost-effectiveness of alternative treatments.
From the outset the Group therefore sought a ‘common core’ of basic information or key attributes to be collected by all investigators in a standardised way. This came to be crystallised in the following set of objectives: (1) To develop a generic instrument to describe and value health-related quality of life (HRQoL), providing both a descriptive profile and an overall index. (2) It was to be a standardised tool to facilitate the collection and pooling of a common data set. (3) It was to be suitable for self-completion and acceptable for use in postal surveys (at that point, a common mode of data collection). These objectives in turn led to a number of requirements for the descriptive system: (1) dimensions should be relevant to patients across the spectrum of health care and to members of the general population. (2) It should be simple—using as few dimensions as possible, with as few levels as possible within each dimension. (3) It should be amenable to self-completion in a range of settings, should be simple enough not to require detailed instructions, and should only take a couple of minutes to complete.
The Group discussed various alternatives with respect to the selection of dimensions, including a survey of patients and the general population, to identify common dimensions of relevance to all groups. Since the selection of dimensions from such an exercise would still involve value judgments, the Group members decided instead to draw on their own expertise by undertaking a detailed review of other available generic health measures. Contrary to expectations, the dimensions suggested for inclusion as a result of this exercise were broadly similar, differing more on dimension nomenclature than on content. General agreement settled on the following: mobility, daily activities and self-care, psychological functioning, social and role performance, and pain or other health problems [4]. In addition, as the Group was multilingual, the classification system descriptors were selected from the outset with a number of languages in mind, rather than a source version being translated into other languages.
The EuroQol Group publicly introduced a six-dimensional health status instrument after some 3 years of development [1]. However, by the time of its publication, further empirical testing had already led the instrument to be further refined to five dimensions—mobility, self-care, usual activities, pain/discomfort and anxiety/depression—each with three levels. Originally named the ‘EuroQol instrument,’ it was formally designated ‘EQ-5D’ in 1995. The descriptive system defined (35) = 243 different states. Two further states were initially included in valuation work: unconscious and dead (both states undefinable in terms of the descriptive system). With the development of the five-level version EQ-5D-5L (see Sect. 2.4), the three-level version was re-designated EQ-3D-3L. (Both versions appear as Appendices 1–4 in the online Supplementary Material).
Initial EQ-5D valuation work employed ranking, magnitude estimation and visual analogue scale (VAS) approaches, but VAS was quickly established to be the valuation approach of choice. At that point in time, other methods were in their infancy, such as time trade-off (TTO), or had not been much applied in the health status context, such as the standard gamble (SG). It was for that reason that the EQ VAS was introduced as part of the EQ-5D questionnaire right from the start: its initial role was actually as a warm-up task for the VAS valuation tasks, and only later was its potential usefulness as a self-reported global measure of overall health recognised [5].
While these early efforts converged on a descriptive system in what was a relatively short period of time, a considerable and rapidly expanding research programme continued, to test the reliability and validity of the EQ-5D in populations and patients. This was accompanied by an extensive programme of research on the valuation of EQ-5D, to test the effect on values of the stated duration of states; the visual presentation and positioning on the VAS scale; the selection of the states to be valued; and deliberation about whose values (experts, patients, or the general public) should be used. It is important to note that these efforts preceded—by over a decade—the establishment of formal health technology assessment (HTA) organisations and processes, so the EuroQol Group was operating in largely uncharted territory.
Consolidation and Expansion
The two and a half decades which followed the establishment of the EQ-5D in 1990 were characterised by continued research and development, considerable growth in the use of the EQ-5D in healthcare decision-making, and ongoing efforts to develop both additional instruments within the EQ-5D framework and improved methodologies for eliciting and modelling health state values. There were also significant changes in the EuroQol Group as an organisation—it grew, formalised its processes and put in place the business model which exists today.
First, it is noteworthy that, apart from some minor wording and design changes to the original EQ-5D questionnaire, what is now termed EQ-5D-3L has remained more or less unchanged from 1990 to the present day. While there has been ongoing experimentation with additional dimensions and the number of levels, as discussed below, these changes have not been incorporated in the EQ-5D-3L instrument itself.
This stability in the EQ-5D 3L instrument has had a number of consequences. After two and a half decades of use and research, there is a substantial back-catalogue of studies, evidence and EQ-5D data available to support new investigations. Research has built upon and developed knowledge of the use and analysis of EQ-5D data. From the perspective of its application in HTA, this stability can facilitate consistent decisions over time.
Expansion in the use of EQ-5D post-1990 came in a number of ways. First, the demand for EQ-5D data and the accompanying value sets increased markedly as HTA organisations became established in healthcare systems around the world. Second, considerable resources were devoted to expanding the number of EQ-5D language versions, facilitating global use of the instrument. Third, there was a rapid increase in the number of applications for licences to use the EQ-5D in a variety of medical and health sector settings, and pharmaceutical companies began to use the instrument in increasing numbers, reflecting the requirement of HTA bodies to supply evidence on QALYs.
In the valuation context a noteworthy development came from the Measurement and Valuation of Health (MVH) study, led from the University of York in the early 1990s, in the form of a set of EQ-5D ‘tariffs’ based on TTO values from the general public which could be used to generate QALYs. The tariff (value set) produced from the MVH study [6] became very widely applied in economic evaluation, both in the UK and in other countries (and continues to be used today). This subsequently led to a number of other countries adopting similar methods for collecting and modelling their own value sets.
Also of significance for the progress of the EuroQol Group was the EQ-net project of 1998–2001 funded by the Biomed programme of the European Commission. This project provided the opportunity to put members’ research work into a more structured context, with most of the efforts of the Group devoted to it during this 3-year period.
The tasks involved were divided into three sub-projects: Translation, Valuation and Application. In addition, the communication of information and knowledge about EQ-5D was addressed, with detail on all aspects of the project subsequently being published in book form [7]. Since the main aim of the project was to harmonise data on the valuation of EQ-5D health states collected in different European countries, considerable effort was put into the Valuation sub-project. Two databases were established, one containing VAS valuations and the other TTO valuations. The Application sub-project produced standard operating procedures (SOPs) for the design, analysis and reporting of EQ-5D in clinical, economic and population studies, which were included in the book alongside guidelines for differing modes of administration of EQ-5D: versions for observer, face-to-face administration, proxy and telephone. The work accomplished in the Translation sub-project is treated separately in Sect. 2.4 below.
Essentially the EQ-net project stimulated the further development and dissemination of EQ-5D, which fed into the scientific programmes pursued in the new millennium.
The Evolution of the EuroQol Group as an Organisation
As use of the instrument grew, the relatively simple club-like nature of the early Group necessarily evolved into more formalised arrangements. The use of EQ-5D in HTA (particularly by NICE in the UK, which, in 2004, identified the EQ-5D as its preferred instrument [8]) led to increased demand from pharmaceutical companies wanting to include EQ-5D data in HTA submissions. This presented an opportunity to license that use and to generate revenue. A key period in the evolution of the Group in this respect was 1993/94. Up until 1993, the activities of the Group were supported exclusively by the initial small group of members and their institutions, both by contributing their time and, occasionally, by contributing financial support to the enterprise.
The next development was the appointment of a Business Manager and a Management Assistant in 1993 and 1994, respectively. In 1994, inquiries from the pharmaceutical industry began to be directed through the business office, which was instructed to develop a pricing policy. This marked the beginning of modest revenue generation, in keeping with the not-for-profit nature of the Group.
This was quickly followed by setting in place legal arrangements. In 1995 the formal organisational (and legal) structure for the Group comprising the EuroQol Association and Foundation, monitored by a Board and Executive Committee, were established under Dutch law, and a Business Management office set up in Rotterdam.
Critical to understanding the current nature of the EuroQol Group was the business model which emerged from this process of formalising the organisation in the mid-1990s. The key features are:
-
The EuroQol Group in all its activities is a not-for-profit organisation.
-
Users must register use of the instrument (copyright to which was first asserted in 1990 and was formally transferred to the Association upon its establishment in 1995).
-
Commercial, for-profit users are changed a licence fee for the use of EQ-5D.
-
Not-for-profit academic users are able to use EQ-5D free of charge.
-
The EuroQol Group comprises both a business unit and an international, multi-disciplinary collaborative network of researchers—the members of the EuroQol Group—who drive forward the science surrounding the EQ-5D.
This combination of arrangements proved an appropriate model for promulgating the use of EQ-5D and generated revenue with which to support and fund research. Group membership expanded—currently at around 80—and has become an international rather than a European network. In addition, with some members having a career-long association with the Group there has been a great deal of continuity of endeavour.
Instrument Development
In Sect. 2.2 we noted that the EQ-5D-3L as an instrument has remained largely unchanged from 1990 until the present. However, there have been important related developments, including many new language versions, newly derived EuroQol Group instruments and systematic approaches to valuation for use in producing value sets. We briefly review the principal developments below, after outlining translation and version management issues.
Translation and Version Management
From the outset English had been used as the working or ‘source’ language for the EQ-5D, and the instrument was simultaneously constructed in Finnish, Dutch, Swedish and Norwegian. Draft translation guidelines were first developed in 1994, and in 1996 expanded guidelines were implemented, overseen by a Translations Committee, setting in place a standard forwards-and-backwards translations process which supported the development of a large number of language versions in subsequent years.
When EQ-5D spread to new languages the process of translation pointed to difficulties in language usage and to differences in the conceptualisation of EQ-5D dimensions and items across countries and languages. This led the Group to consider more closely the meanings of concepts and the related wording used in EQ-5D, not least in English. The Translation aspect of the EQ-net project provided the opportunity for substantive work on these matters. A definition of EQ-5D concepts was provided and a series of recommendations for further research was made [9]. Also generated were a taxonomy of definitions of EQ-5D concepts, SOPs and detailed translation guidelines. A detailed account of translating EQ-5D into 11 European languages provided an insight into the translation process, and the challenges involved [10].
In 2009 a Version Management Group (subsequently Committee) was established, with responsibilities for reviewing new language versions, responding to client and translation agency queries, updating essential documentation, and implementing systems aimed at improving version control and management. This group has responsibility not just for different language versions, but also for testing and approving electronic versions of the EQ-5D (tablet, web-based, PDAs), demand for which has risen [11].
EQ-5D-5L
Notwithstanding the strong uptake in the use of EQ-5D, particularly in HTA, concerns about its adequacy as a measure of HRQoL have been voiced. There continued to be lively debate within the EuroQol Group going back to 1994 [12] regarding the three-level structure of the response options (no, some, extreme problems/unable to) and whether this was associated with ceiling effects and a lack of sensitivity to changes in health. Kind and Macran fuelled that debate, reporting an investigation of a five-level version of the core five dimensions [13]. In 2005, sufficient momentum on this issue had built such that a EuroQol Group task force was established to consider an increased level descriptive system, in response to concerns about the perceived lack of sensitivity of the EQ-5D and ceiling effects in the descriptive system.
In 2006, after considerable debate and pre-studies about whether to go for a four- or five-level version, it was decided to recommend the development of a five-level version of the instrument, while retaining the same core five dimensions. In addition ‘confined to bed’ was replaced by ‘unable to walk about’ to increase sensitivity of the mobility dimension. Results from initial studies testing five-level versions of the EQ-5D showed increased reliability, sensitivity (discriminatory power) and feasibility [14–17].
As with the original EQ-5D, the intention behind the new five-level version was that it be accompanied by value sets. For that reason, rather than assign the additional two levels as ‘unlabelled’ response options between no and some, and some and extreme, problems, it was felt that all five levels required labels. Labels were selected following the results of semantic testing in England, Spain and France [18, 19], and the resulting labels translated into other languages. Two features of this process can be highlighted. First, the labels were chosen for UK English, Spanish and French, based on an exhaustive process of response scaling among a wide range of potential labels selected from the literature and existing PRO questionnaires, together with follow-up focus group research to explore respondents’ understanding of those labels in the three countries. Second, once the labels had been decided on, the UK English, Spanish and French versions could then be used as source content for any new language versions required. These are produced following the EuroQol Group’s translation methodology, which also includes in-depth semantic testing of all wording in the target language, with a particular focus on the severity labels (see [11]).
The new instrument was approved as an official EuroQol instrument in 2009. From that point, the five-level instrument has been referred to as the EQ-5D-5L, and the original EQ-5D has been re-named the EQ-5D-3L. The research underpinning the EQ-5D-5L is summarised in Herdman et al. [20], including consideration of the mapping of health states from one system to the other.
EQ-5D-Y
The EQ-5D was, implicitly, designed for self-completion by adults. However, HTA bodies and other healthcare decision-makers frequently make decisions regarding treatments for children and young people. From 1998, interest grew in the possibility of using the EQ-5D, or adapting it in some way, for use in younger people. An initial ‘child friendly’ version of the EQ-5D, reported in 2002 [21], was followed by research efforts in a range of countries. A task force, established in 2006, coordinated these efforts, and considered issues regarding what dimensions to include and how to label them, what number of levels to use, proxy completion, what age ranges to target, and how to value children’s health states [22, 23].
These efforts culminated, in 2009, in approval of a ‘youth’ version, the EQ-5D-Y, as an official EQ-5D product. The EQ-5D-Y retained the same five-dimension, three-level format of the EQ-5D, but dimensions were described in more appropriate language as: mobility (walking about); looking after myself; doing usual activities; having pain or discomfort; feeling worried, sad or unhappy. (EQ-5D-Y appears as Appendix 5 in the online Supplementary Material).
EQ-5D-Y is suitable for self-completion by children aged 8–11 years; it is also recommended for use at ages 12–15 years, although use of the EQ-5D adult version might be possible in some circumstances. The EQ-5D adult version is recommended for those aged 16 years and over.
Protocols for Value Sets
Despite the widespread utilisation of the MVH tariff for QALY purposes there was no ‘official’ valuation protocol or consensus view within the EuroQol Group about valuation methods to be used in producing value sets. Different research teams adapted the MVH study design in various ways, making different choices about, for example, the number and selection of states to value; ‘exclusion rules’ applied to the data; and so on [24]. This limited the comparability of the data. This was addressed in 2009 at a meeting in Paris, where a modified version of the MVH study design was endorsed (‘the Paris protocol’) for use in EQ-5D-3L value set studies. Included among the changes incorporated at that point was dropping ‘unconscious’ from the states to be valued in such studies.
Having developed the EQ-5D-5L, the EuroQol Group decided this presented an important opportunity to improve valuation methods and to promote a consistent approach to valuing EQ-5D-5L by providing an official protocol and study design. Interim values for the EQ-5D-5L were available from a ‘crosswalk’ study: six countries administered both the five-level and three-level versions in parallel, from which a crosswalk enabled EQ-5D-5L profiles to be mapped to EQ-5D-3L profiles, and values applied from existing EQ-5D-3L value sets [25]. At the same time, a series of methodological studies were undertaken, exploring a variety of approaches to both TTO and discrete choice experiments (DCE). The latter method had been investigated in 2008 using the three-level version [26]. Work was also undertaken to develop thesoftware to allow these methods to be implemented in computer-assisted personal interviews (CAPI). A prototype protocol, incorporating these approaches, was piloted in a multi-country study [27].
Following further testing and refinement, the international protocol for valuation of EQ-5D-5L was launched [28]. This comprises a fully documented study design, interviewer guide, interviewer training materials and the CAPI software, EQ-VT (EuroQol Valuation Technology). Valuation studies in England, The Netherlands, Spain and Canada commenced in 2012, coordinated and supported by the EuroQol Executive Office.