Submission from David Parkin
There’s a problem with the way that health economists and others describe the properties that a health state index should have. The main reason we want such an index is to calculate Quality Adjusted Life Years. So, we need the index’s possible values to run from zero to one, though we can also tolerate negative values. But what does zero mean in this context?
There aren’t too many problems with saying what we mean by a health state that has a value of 1. It’s described using terms like ‘full health’. That’s interpreted to mean that one year spent in ‘full health’ will generate one Quality Adjusted Life Year. 1 QALY is as much health as any one person can have in a year. There’s room for debate about what ‘full health’ means, in particular its subjective interpretation, but this is a detail about a coherent concept based on the idea of what a QALY is.
But it’s much more difficult to define what the value 0 means. Health economics texts often define it as ‘dead’ or even ‘death’. This is then followed by an explanation of what negative values mean, leading to the concept of ‘worse than dead’ or ‘worse than death’. I think that this definition is wrong. It’s misleading and may bias the results of health state valuation studies.
The most usual applications of the QALY model don’t aim to compare health states among the living with ‘being dead’. They compare different health states amongst the living. If 0 and 1 are intended to be health state values, they should be defined with respect to health states. ‘Full health’ is a health state, but ‘being dead’ is not, except perhaps to zombies and vampires. In fact, if you rate a dead person in EQ-5D health state terms, they will be a 33311. Unable to do anything, but in no pain and not anxious or depressed.
Of course, it’s important that ‘dead’ is valued at 0. ‘Full health’ must be valued at 1, because an essential QALY property is that every year of life spent in full health produces 1 QALY. Similarly, ‘being dead’ must be valued at 0 because another essential QALY property is that dead people produce no QALYs. But are there are other ways of producing no QALYs?
No QALYs are produced if there are no life years, but being dead is not even the only way to achieve that. It can also be achieved by not being alive in the first place. More importantly, a living person will also enjoy no QALYs in any year that they spend in health states that are valued at 0. In the same way that the health state value 1 implies as much health that we can have at a given time, the health state value 0 implies a complete absence of health at a given time. But what does that mean? They are not the worst health states that people can have, since to some people those worse states generate negative values. So, what do health states with a value of 0 look like?
One way to solve this problem is to observe that they are health states that are as bad as being dead. That gives us a way of thinking about them and of describing them to people for valuation studies. This has an additional advantage. It is more consistent with the idea of negative health values than is the idea that being dead defines zero. Negative values refer to levels of healthiness not deadness. It’s OK to describe health states with negative values as being worse than dead. But that isn’t their essential feature, which is that they generate negative QALY values over time. Negative numbers mean a very bad health state, not an extremely dead state.
Health state valuation studies in general use the term ‘dead’ for comparison with health states, explicitly or implicitly meaning 0. But when being asked to value health states, can people really imagine ‘being dead’ in any meaningful way? Some studies also use an even worse term, ‘death’. The only route to being dead is to die. But dead, dying and death are not the same things. Dead is a state, though not a health state, but dying is a process and death an event. Death and dying may be valued as negative in welfare terms even if dead is correctly valued at 0. Using death as a comparison when valuing health states is likely to distort the values obtained.
Of course, many people recognise the issue that I’ve raised, but presumably they think that it is not important. I think that it’s worth researching whether or not it matters empirically. If not, I guess it’s a dead issue.
[…] with the way we handle ‘dead’ in health state valuation. I think it’s a more serious issue than we know (and we know quite a bit), so I am always glad to see attempts to fix it. Once you get […]
Sorry, didn’t realise this post originated in 2011 so health economists have not been a slow off the mark as I implied in my first comment
Glad to see this issue finally getting discussed….I discussed it in 2008 in my BMC paper 🙂
That would be “Rescaling quality of life values from discrete choice experiments
for use as QALYs: a cautionary tale” (Flynn et al 2008?) available here http://eprints.bham.ac.uk/141/1/coast_pop.pdf
Surely all you’re really doing is quibbling about whether the EQ-5D scale should start at -0.59 or at zero. If you were to consider evaluating health-related quality of life interventions with suicidal populations it would make perfect sense for death to be valued somewhere within the range of possible health states. It just so happens that it’s 3/8 of the way along the scale, which may be surprising but is also eminently plausible. Else how could you model an individual’s rational choice of death over life?!
I like the 33311 quip but really these people are not ‘zombies and vampires’ and there is the usual danger here of invalidating a broad range of severe and enduring mental health conditions – and in any case how exactly do you assert that a reanimated corpse would score 11 on pain and anxiety?! I suppose the qualitative distinction you make between the health-related benefits of death and being dead does have some validity in the case of a martyr or suicide bomber … but even such a person is engaged in a form of TTO analysis albeit somewhat eccentric.
[…] are many problems with the way we value death; some have been discussed on this blog. One which has become clear to me recently is that we don’t seem to fully take […]
[…] so much as ‘death’ is used in the valuation of health states. This is problematic, as has been discussed. Economists haven’t really gotten round to disagreeing about this yet, but there’s […]
Interesting paper in Quality of Life Research looks at the extent to which a person’s valuation of euthanasia is reflected in their TTO valuations: http://www.ncbi.nlm.nih.gov/pubmed/22678351
[…] What does a Health Value of Zero Mean? « The Academic Health … […]
Chris, do people get any real utility purely from being alive though? If this were the case, there would be no demand for assisted suicide. By setting 0=dead we are implying that people are in such a bad state that they cannot look past their condition to gain utility from the other ‘lifestyle’ or ‘social’ factors you mention. I think that there must be cases where this occurs and people feel they are gaining no utility just through the process of living. Thus the ability to value health states as worse than being dead is relevant and important to real world applications.
An anchor at worst health state imaginable is possible but would be complicated to quantify because what people consider the worst possible state is highly objective. When I researched why people value health states as worse than being dead under time trade off, there was no significance found for the order in which respondents were presented health states and whether an individual health state came before or after the worst state (33333). However, there were statistically significant relationships seen between education levels, socio-economic background and employment (amongst other factors). As such when creating an anchor at the worst possible health state which could be used across the general population, how could these factors be accounted for?
I’m not saying people gain utility from simply being alive (though I expect many people would say they do), I’m saying that people gain utility from things in their life aside from their health, so putting a ‘dead’ state on a health state valuation makes no sense. To be dead is to have an absence of life, not just an absence of health. I think if you look at the EQ-5D valuation questionnaires it isn’t specified that zero means “a state where you are in such a bad state that you cannot look past your condition to gain utility from anything else”. That would be a methodological nightmare.
I have no doubt that, in some cases, people would rather be dead than alive, and that health would be an important consideration in this. The point is that it is not the only consideration and as such a health state valuation has no right to incorporate ‘dead’ on its scale. If there’s one thing nobody knows how to value it’s being dead; there’s no way a valuation of it could be objective and I expect a valuation of ‘worst state imaginable’ would be more realistic. Likewise, the health state at which an individual would prefer to be dead is massively subjective. I think in a valuation of a ‘worst health state imaginable’ things would be controlled for in the same way as they are for valuations of ‘deadness’ – not at all.
Personally I think questions of life and death such as this fall beyond the remit of us mere health economists, and I don’t think they should be our priority. So long as we are aiming to maximise an individual’s health, whether they would or would not rather be dead is somewhat inconsequential so long as we know their health state (and life-years saved is usually collected separately anyway). Valuing health should be our priority, but right now we’re not getting it right.
The way I tend to think about it is mostly in terms of how a dead person moves through a Markov model or decision tree: once a person is dead they are effectively removed from the model. They no longer generate any QALYs, but they also no longer generate any costs and they cannot move into another health state i.e. come alive again.
What I find a bigger problem is complex conditions that are high cost if the person survives such as heart attacks and stroke. If you keep more people alive they go on generating costs, where as the dead people don’t. Clinicians for this patient group will usually use 30 day mortality as an anchor for the effectiveness of related interventions, where as we tend to use lifetime costs and QALYs. Try telling a clinician that an intervention isn’t cost effective if their 30 day mortality rate is significantly lower!
This is a good point. I’d say clinicians care about mortality in this case because that’s what matters to the patients. Their valuation of their own mortality isn’t based on health alone, I suspect it never is – other stuff matters (love, responsibility, a hard-earned pension). However, by setting 0=dead we are implying that there is only value in life if somebody has some positive level of health. A person may very well value their health state to be equal or worse to the health state that they would assign to being dead… But it is by no means logical to think that this means they would rather die. Basically, it is possible to value health as less than zero whilst valuing life as greater than zero.
This=problem
I think this is an issue that really needs more discussion; the ‘dead’ anchor causes a lot of problems in health valuation. I think this issue actually highlights the key problem with QALYs: that definitions of ‘quality’ and ‘quantity’ of life don’t sit as comfortably together as we’d like.
Do you believe we could completely do-away with the ‘dead’ aspect? Possibly switching to an anchor at ‘worst health state imaginable’? It may be far from objective, but it seems more open to valuation than being ‘dead’.