Where’s WALY? : A proof of concept study of the ‘wellbeing adjusted life year’ using secondary analysis of cross-sectional survey data

When making decisions about interventions it is important to consider their effect on both length of life and the quality of that life. One way in which this is achieved is through the use of Quality-Adjusted Life Years (QALY). The most common method of calculating QALYs uses a measure called EQ-5D-3L [1, 2] which has been successfully employed to assess the relative effectiveness of a wide range of treatments and interventions. However, there is evidence of ceiling effects in the EQ-5D-3L, with up to 85 % of respondents who have physical health problems reporting maximum scores [3, 4]. Further, there are questions about whether the EQ-5D-3L is appropriate for assessing the impacts of conditions such as hearing loss, age-related macular degeneration, diabetic retinopathy and psychotic disorders. This is because of problems such as failure to detect differences in quality of life between people with different stages of disease severity, and a limited ability to detect improvements in quality of life following interventions [4–9].

Wellbeing is now recognised as a determinant of longevity and an important player in the adoption and maintenance of healthy lifestyles and successful management of chronic illness [10]. The case for improving wellbeing has been made on both health and economic grounds [11]. Interventions to promote mental wellbeing (e.g. parks and gardens, crime reduction, art festivals, cookery clubs, wellbeing festivals, Tai Chi, yoga, sports) may be offered in many different sectors, both public and private and it is important to be able to assess their relative effectiveness compared to interventions offered in the health sector. While researchers have mapped utility of the EQ-5D-3L onto utilities derived from a range of health outcome measures (e.g. SF-6D) [12–14], there has been little research on how to address the cost-utility of interventions aimed at improving mental wellbeing. If wellbeing is a concept that substantially extends existing concepts of health, then a health-related measure of quality-of-life will underestimate the benefit of interventions that improve wellbeing. In a time of austerity, this is clearly an issue for public health commissioning. One approach to address the cost-utility of these types of interventions could be to develop a wellbeing adjusted life year (WALY).

A well-established tool to measure mental wellbeing is the Warwick-Edinburgh Mental Well-being Scale (WEMWBS) [15, 16]. Full details of the WEMWBS are available (www2.warwick.ac.uk/fac/med/research/platform/wemwbs/). In brief, WEMWBS was developed to meet the need for a robust, population-based measure of mental wellbeing to evaluate programmes and monitor mental wellbeing at the population level [15]. WEMWBS has been in use since 2007, it is valid and reliable at the population level [15] and is sensitive to change [17]. Originally validated in English and Scottish populations of people aged 16 and older [15], the scale has now been translated into many different languages and validated in many different cultures [18]. It has been successfully used to measure wellbeing outcomes in a range of health interventions [19–25]. Evidence suggests that users of mental health services and their carers prefer the WEMWBS to other health outcome measures [26].

WEMWBS is gaining momentum as a useful tool in public health practice, particularly since its inclusion as a measure in the Scottish Governments Outcomes Framework [27] and the English Public Health Outcomes Framework [28]. However, there is no underlying research on how to assess the cost-effectiveness of interventions using this tool. In this study we explore the extent to which the WEMWBS and the EQ-5D-3L estimate the health state value of individuals with different levels of mental wellbeing (i.e. whether WEMWBS “maps” onto the EQ-5D-3L) [29]. If mapping is poor, this suggests that there is a need to develop a Wellbeing Adjusted Life Year (WALY).

The aim of this study was to assess the extent to which WEMWBS maps onto the EQ-5D-3L. Consistent with previous studies [14], we found a pronounced ceiling effect in the EQ-5D-3L, with nearly three quarters of participants having the maximum score of 1 (i.e. the best possible health-related quality of life). No ceiling effect was observed for WEMWBS. WEMWBS scores spanned the whole range of possible values (14–70), with a mean of 53.9 for participants who had a score of 1 on the EQ-5D-3L. This suggests that WEMWBS and the EQ-5D-3L are not measuring the same construct and that there is scope for improving mental wellbeing of individuals who have maximum scores on this quality of life measure. WEMWBS was positively correlated with both the EQ-5D-3L and the EQ-5D-3L VAS for the sample as a whole and when stratified by age, sex, and socioeconomic status, though this correlation was quite low. Both WEMWBS and the EQ-5D-3L detected differences between those with very good versus other levels of self-reported health, but neither measure was especially good at detecting these differences. This is not surprising as the constructs that they are measuring are not identical. WEMWBS explained a very limited amount of the variability of the EQ-5D-3L and the EQ-5D-3L could not assess with any precision the effectiveness of interventions to promote mental wellbeing relative to other health related interventions.

It is essential that the preference-based measure adopted by an economic evaluation captures all consequences of the alternatives being evaluated that might materially affect the net benefit of each alternative to the decision-maker. It can be argued that wellbeing is a concept that extends existing concepts of health, prompting the need for a health-related measure of quality-of-life that will not underestimate the benefit of interventions that improve wellbeing. This remains an issue for interventions in sectors such as social care and education, and is increasingly relevant for public health and mental health interventions [27, 28]. The EQ-5D-3L has been shown to capture the impact of health care interventions for a broad range of conditions, but the fact that we found a ceiling effect in the EQ-5D-3L (as have others before us [51], with nearly three quarters of participants at the maximum score reinforces the likelihood that it does not capture relevant changes that matter to individuals or, therefore, to economic evaluations [51]. The EQ-5D-3L is preference-based, i.e. tariffs exist that reflect societal preferences for different types of health gain, relative to life extension, to permit calculation of QALYs. If a similar tariff existed for WEMWBS that allowed estimation of Wellbeing Adjusted Life Year (WALYs) gained, this could be used to support priority-setting within and across sectors in a way that reflects societal preferences more appropriately. Further research would be required to understand how the overlap between health and wellbeing varies in different populations, and to determine how the QALY and/or WALY can be used to value the benefits of interventions in these populations, while avoiding double-counting.

A limitation of our study is that we compared WEMWBS to the EQ-5D-3L. A new version (EQ-5D-5L) has recently been published [52] which has ameliorated some of the limitations of EQ-5D-3L discussed in this paper and has reduced ceiling effects with increased discriminatory power [53].

WEMWBS has the potential to be used as the basis of a preference-based measure to evaluate and prioritise public sector interventions between and within sectors, including traditional health related interventions. However, it has not yet been used to inform priority-setting and a preference-based tariff does not currently exist. The next step for our research is to develop and evaluate the utility of a well-being adjusted life year (WALY) based on WEMWBS. The stages in the development of the WALY will include a valuation exercise to generate a preference tariff for WEMWBS, the identification of an appropriate preference elicitation technique for wellbeing states, and exploration of the variation in valuations across samples.

There is wide variation in the levels of mental wellbeing amongst participants with very high levels of self-reported health who score at ceiling level on the EQ-5D-3L with evidence of limited mapping of WEMWBS onto EQ-5D-3L. These results suggest that the two measures examine related, but not identical, aspects of quality of life. We propose exploration of the feasibility, appropriateness, and practicality of a Wellbeing-Adjusted Life Year.