Back in 2005, NICE and the Department of Health here in the UK set about trying to figure out the level at which society valued a QALY, and also whether different QALYs should be assigned different values. The results of the Social Value of a QALY (SVQ) project have recently been published, and last month the authors involved presented a summary of their findings along with their own interpretation of their results. A crucial read.
The SVQ project is certainly a noble cause and should be a key ambition for health economists to pursue. The authors of the project, in their recent paper, highlight a number of problems and controversies in their methodology. The next step is clearly to investigate these and other methods further before embarking upon a QALY valuation project of a greater magnitude. While the authors recognise the need to hone these methods, it seems to me that a number of other methodological problems have been missed. I would like to raise one here.
Mr QALY: “dead=zero” (Mr VPF: “dead=…..”)
The first part of the study sought to elicit society’s valuation of a QALY using VPF (value of preventing a fatality) and VSI (value of serious injury) figures, as used by the Department of Transport. The results showed that people valued different ‘types’ of QALY differently; that is, QALYs calculated based on life-saving, life-extending and quality-of-life-enhancing considerations. These were valued at £70,000, £35,000 and £10,000 per QALY respectively.
However, there is another way to think about this. It might be that, in reality, people actually value these different types of QALY equally. If this were the case then this discrepancy may arise because VPF respondents actually value being dead as being ‘less than zero’. At first glance this doesn’t make sense as, in the original valuation of QALYs, being dead was anchored at zero. However, to my knowledge, such a valuation of being dead (even if indirect) was not included in the calculation of the VPF. The result of this, in my opinion, is that individuals have to assign their own value to being dead, as it is not pre-determined for them.
It seems possible that, in the calculation of the VPF, individuals assigned an extra negative value to the act of being killed in an accident. This may be caused for a number of reasons – not least an individual’s immediate emotional reaction to a question about, for example, being killed in a road accident. It also seems possible that this is not simply a bias in the method of collecting data but a real dynamic in the overall effect of someone being killed in an accident. In essence, the VPF measures society’s view of the effect of an individual losing their life in an accident, and not the effect of an individual moving from 1 to 0 on an EQ-5D scale. This would mean that a year in “full health” in the QALY calculation is measured on a different scale to a year in “full health” in the VPF calculation. ie – death could result in a greater perceived ‘loss’ in the VPF than the perceived ‘change’ form 1 to 0 in the EQ-5D.
Formulaically I see the problem being that:
VPF=WTP(Y,L)
where WTP(Y) is the monetary value that an individual assigns to each life year lost(quality-adjusted or otherwise), and WTP(L) is a fixed value assigned to the fact that an individual is losing their life. The SVQ study then calculates the value of a QALY from this as:
VQ=VPF/R
where VQ is the value of a QALY and R is the expected number of remaining life years for that individual. This seems reasonable.
However, the VSI would not include this extra ‘L’ factor and thus the WTP will be lower, even if the QALY loss is equivalent. While the EQ-5D is not combined with the VPF data, it is later combined with VSI data, giving the figures for the value of different QALYs shown above. To me this makes the figures entirely incomparable as quality-of-life-enhancing QALYs are based on EQ-5D valuations, while life-saving QALYs are not.
For this reason I feel the valuations of different ‘types’ of QALYs needs to be taken with a pinch of salt. It seems to me that the VPF and VSI should not be used in trying to value the QALY in the future, as we are not comparing like-with-like.
Please share your thoughts on this matter in the comments box below, and please note that this is all conjecture and I may very well be wrong!
