There are three EQ-5D value sets available for use in cost effectiveness analysis in the UK and/or England:
– the UK EQ-5D-3L value set (often called the ‘MVH’ value set) which has been used for two decades, e.g. in evidence submitted to NICE (Dolan, 1997);
– a ‘crosswalk’ value set (van Hout et al., 2012), which maps patients’ EQ-5D-5L data to the EQ-5D-3L descriptive system, so that the MVH value set can be applied; and
– the EQ-5D-5L value set for England (Devlin et al., 2016).
How do these value sets differ? And how important are these differences likely to be for users?
A new OHE Research Paper by Mulhern et al. (2017) compares the characteristics of the value sets.
Written by Nancy Devlin, Koonal Shah & Brendan Mulhern
There are three EQ-5D value sets available for use in cost effectiveness analysis in the UK and/or England:
- the UK EQ-5D-3L value set (often called the ‘MVH’ value set) which has been used for two decades, e.g. in evidence submitted to NICE (Dolan, 1997);
- a ‘crosswalk’ value set (van Hout et al., 2012), which maps patients’ EQ-5D-5L data to the EQ-5D-3L descriptive system, so that the MVH value set can be applied; and
- the EQ-5D-5L value set for England (Devlin et al., 2016).
How do these value sets differ? And how important are these differences likely to be for users?
A new OHE Research Paper by Mulhern et al. (2017) compares the characteristics of the value sets.
What we found:
The paper compares the distribution of predicted values in each of the three value sets – that is, the MVH 3L values for all 243 states described by the 3L instrument, and the crosswalk and 5L value set values for all 3,125 states described by the 5L instrument. The overall range of values (the difference between the minimum and maximum values) is smaller in the 5L value set than in the MVH 3L or crosswalk value sets. The change in values between adjacent states (that is, where there is a one-level change on one dimension) is also smaller on the 5L than on the 3L.
We looked at the values of ‘matched’ states – for example, the ‘middle’ state on the 3L (state 22222, describing some problems on all dimensions) compared with the ‘middle’ state on the 5L (state 33333, describing moderate problems on all dimensions). 5L values were higher for all matched states.
We also applied the value sets to 3L and 5L data from patients who had completed both measures. The 5L value set produces higher values across all of the conditions included, and the differences are generally significant. However, there is some evidence that, despite these differences, the value sets broadly rank different health conditions in a similar way, in terms of their severity.
Are these differences surprising?
The results were pretty much what we expected, for two reasons.
First, the 5L has more levels with which a patient can self-report health problems. The ‘jumps’ in severity between levels are therefore smaller in the 5L than the 3L, by design. So we would expect the differences in utility between adjacent states to be smaller – which is exactly what we found.
Second, the methods used in the 5L value set addressed problems known to affect the MVH 3L value set. As we have pointed out (Devlin et al., 2010, Devlin et al., 2013), the MVH 3L value set is a bit unusual in that it has a relatively low minimum value and a large number of states with negative values – these are states that are considered ‘worse than dead’. Nearly a third of the values in the MVH 3L value set are negative, compared to about 5% in the 5L value set. The international protocol used to value the 5L (Oppe et al., 2014) yields a minimum value of -1, compared with a minimum value of -39 in the protocol used in the MVH 3L study (values from which were rescaled for analysis). It was therefore entirely expected that any new value set for the 5L in England would be likely to have higher values than the UK MVH 3L value set.
What will these differences mean for users and decision-makers?
Although our new OHE Research Paper shows that the UK MVH 3L and England 5L value sets have very different characteristics, the overall impact of that on estimates of quality-adjusted life-year (QALY) gains and incremental cost effectiveness ratios (ICERs) is not straightforward to assess. For example,
- The 5L value set has a narrower range thus higher values, but using the 5L descriptive system is likely to detect smaller changes in health because it has more levels (Feng et al., 2015).
- The overall effect is a combination of both the greater sensitivity of the 5L descriptive system and the higher values of the 5L value set, compared to their 3L counterparts.
- The impact of using the 5L rather than the 3L on estimates of QALYs gained is unclear as they will apply to both control and intervention groups, and will depend on whether the gain is in quality of life, survival, or a mix of both.
Some results on this have recently been published by the NICE DSU. The DSU paper uses a mapping approach to translate 3L to 5L data and vice versa. Their results suggest that the differences in values noted in our OHE Research Paper will in most cases result in reduced estimates of QALY gains and less favourable ICERs. The opposite was the case in situations where there were gains in length of life (because of higher 5L values applied to added years). However, there are some limitations to the DSU study, e.g. the mapping results differed depending on which patients’ data were used.
What remains to be known?
Data from patients asked to complete both the 3L and the 5L, at various time points, are required to better understand the differences between the instruments in measuring self-reported health.
The differences between the UK MVH 3L and England 5L value sets arise in part because of improvements in methods used in 5L value set studies. This might suggest a need to update 3L value sets, using similar methods to those used in 5L value set studies. Another reason for doing that is that the MVH 3L value set is now over two decades old – and there are likely to have been changes in the preferences of people and in the demographic composition of the population over time.
Given the importance of these issues for users and decision makers, a EuroQol Group Taskforce, co-chaired by Professor John Brazier and Professor Nancy Devlin, has been established to take research comparing the 3L and 5L forward.
FAQs
1. Is the Devlin et al. (2016) value set ‘official’?
The EQ-5D-5L value set for England reported by Devlin et al. (2016) and analysed in Mulhern et al. (2017) and Wailoo et al. (2017) remains provisional at the time of writing, as the paper is under peer review with a journal. It should be considered as having interim status as the peer review process may necessitate changes to the analyses and results. When the paper is accepted by a journal, the value set reported in the journal article will supersede any previous versions. Note that the EuroQol Group does not offer any formal ‘official’ vs ‘non-official’ designation for any value set. NICE has yet formally to consider how the availability of a 5L value set will be reflected in future revisions to its current methods guide.
2. Is the new 5L value set for England or for the UK?
While the MVH 3L value set was based on the stated preferences of the general public of the UK, the 5L value set reported in Devlin et al (2016) was based on the stated preferences of the general public in England only. This is because the study was funded by the English Department of Health. However, once we have confirmation that our value set paper has been accepted by a journal, we will proceed to combine the English data with additional data we have collected in Scotland, Wales and Northern Ireland to also create a UK 5L value set. So at present a 5L value set for the UK is not available. Users of value sets should consider which value set is most appropriate for their study and their needs.
3. What work is OHE currently doing to further investigate these issues?
The 5L value set study used both improved ways of eliciting preferences and improved ways of modelling valuation data, compared to 3L valuation studies. Yan Feng and Ben van Hout are currently working on a new project to apply the 5L valuation modelling methods (Feng et al., 2016) to the original data from the MVH 3L study.
References
Devlin, N., Buckingham, K., Shah, K., Tsuchiya, A., Tilling, C., Wilkinson, G. and van Hout, B., 2013. A comparison of alternative variants of the lead and lag time TTO. Health Economics, 22(5), pp.517-532.
Devlin, N., Shah, K., Feng, Y., Mulhern, B. and van Hout, B., 2016. Valuing health-related quality of life: an EQ-5D-5L value set for England. OHE Research Paper. London: Office of Health Economics.
Devlin, N., Tsuchiya, A,, Buckingham, K. and Tilling C., 2010. A uniform time trade off method for states better and worse than dead: feasibility study of the ‘lead time’ approach. Health Economics, 20(3), pp.348-361.
Dolan, P. (1997). Modeling valuations for EuroQol health states. Medical care, 35(11), 1095-1108.
Feng, Y., Devlin, N. and Herdman, M., 2015. Assessing the health of the general population in England: how do the three-and five-level versions of EQ-5D compare? Health and Quality of Life Outcomes, 13(1).
Feng, Y., Devlin, N., Shah, K., Mulhern, B. and van Hout, B., 2016. New Methods for Modelling EQ-5D-5L Value Sets: An Application to English Data. OHE Research Paper. London: Office of Health Economics.
Wailoo, A., Hernandez Alava, M., Grimm, S., Pudney, S., Gomes, M., Sadique, Z., Meads, D., O’Dwyer, J., Barton, G. and Irvine, L., 2017. Comparing the EQ-5D-3L and 5L. What are the implications for cost-effectiveness estimates? Report by the Decision Support Unit, ScHARR, University of Sheffield.
Mulhern, B., Feng, Y., Shah, K., van Hout, B., Janssen, B., Herdman, M. and Devlin, N., 2017. Comparing the UK EQ-5D-3L and the English EQ-5D-5L Value Sets. OHE Research Paper. London: Office of Health Economics.
Oppe, M., Devlin, N., van Hout, B., Krabbe, P.F.M. and de Charro, F., 2014. A programme of methodological research to arrive at the new international EQ-5D-5L valuation protocol. Value in Health, 17(4), pp 445-453.
van Hout, B., Janssen, M.F., Feng, Y.S., Kohlmann, T., Busschbach, J., Golicki, D., Lloyd, A., Scalone, L., Kind, P. and Pickard, A.S., 2012. Interim scoring for the EQ-5D-5L: mapping the EQ-5D-5L to EQ-5D-3L value sets. Value in Health, 15(5), pp.708-715.
