Decisions, QALYs and the value of a life

Here’s a well known thought experiment:

A runaway train is on course to collide with and kill five people who are stuck at a crossing down the track. There is a railway point and you can pull a lever to reroute the train to a siding track, bypassing the people stuck at the crossing but killing 2 siding workers.

What would you do? What is the ethical thing to do? Why? What if one of the siding workers was related to you? What if the people stuck at the crossing were convicted murderers on their way to prison? What if the people on the crossing were not killed but permanently maimed?

Have a think about it before reading on.

Unless you answered that you did not accept the situation at face value (like Captain Kirk and the Kobayashi Maru simulation), or refused to choose, you will have made some judgements about the relative value of the choices on offer and perhaps the relative value of the lives at risk. You are not alone in this: in a 2018 Nature paper, nearly 40 million people from 233 countries were prepared to make similar choices. On average there were preferences to save the young over the old, the many over the few and the lawful over the unlawful, though with some interesting regional and cultural variations.

Making value judgements about people in a thought experiment is one thing, but making them in the real world with impacts on real people and their lives is another. Ascribing value to a persons life has grim historical and moral connotations. If someone is deemed somehow less valuable than someone else there is a risk that this is used to justify stigmatisation, discrimination, persecution and even genocide. We therefore need to be extremely careful about the moral context in which such judgements are made and the language we use to discuss them. Human rights, justice and the fundamental equivalence of the life and interests of different people must be central. 

Decisions which affect the health, livelihoods and welfare of citizens are (and need to be) made all the time. In some cases decisions affect length or quality of life, or liberty.  Decision making during the pandemic (whether locking down, opening up, isolation, mask wearing or travel restriction) is a potent recent example. Few people would argue that no decisions were necessary even if they may disagree with the details of some (or all) of the actual decisions made.

But if everyone’s life and interests are equivalent, how do we avoid becoming paralysed when faced with choices which inevitably will have (sometimes significant) consequences for different individuals? We do this by understanding that the values we ascribe to the people affected by a decision are not absolute measures of their worth, but merely tokens which allow us to undertake some accounting. If the process by which we allocate these tokens is transparent, just and humane then their use to inform a decision is morally defensible. Choosing to switch the points because this results in the least worst outcome on average is morally very different from choosing to switch them because you have a seething hatred of railway engineers.

What tokens can we use in healthcare to help us make decisions?

There have been attempts to provide a quantitative framework for measuring health. The most commonly recognised token of health is the Quality Adjusted Life Year (QALY) though there are others (eg. Disability Adjusted Life Years [DALY]). One QALY is a year lived in full health. A year lived in in less than full health results in less than one QALY, as does less than one year lived in full health. How much we scale a QALY for less than full health is determined by studies asking members of the public to imagine themselves ill or disabled and then enquiring (for example) how much length of life they’d trade to be restored to full health (time trade-off) or what risk of death they’d accept for a hazardous cure (standard gamble). 

The QALY is a crude and clumsy tool. It has been criticised for relying on functional descriptions of health states (like pain, mobility and self care) rather than manifestations of human thriving (stability, attachment, autonomy, enjoyment), for systematically biasing against the elderly or the disabled and for failing to take into account that health gains to the already healthy may be valued less than health gains to the already unhealthy (prospect theory). The scalar quantities contributing to a QALY (‘utilities’) reflect the perceptions of those surveyed during QALY development, validation and revalidation. These perceptions may be clouded by fear or ignorance and may have little relation to the real experiences of people living with a health impairment or handicap. Some have argued that QALYs have poor internal validity and are therefore a spurious measure.

These are important, though arguable, technical criticisms and to some extent explain the marked international variation in the use of the QALY: they are used in the UK and some Commonwealth countries, but have been rejected as a basis for health technology assessment in the US and Germany. And yet, decisions need to be made. If not QALYs then what else?

But the most emotionally charged criticism of QALYs is that they somehow inherently rank people’s value according to how healthy they are or that the health of people who gain fewer QALYs from an intervention is somehow worth less than the health of those who gain more. This is a misunderstanding. QALYs (like the assessments you made of the lives at risk from the runaway train) are accounting tokens. When fairly, justly and transparently allocated (and technical criticisms might be important here) they merely allow a quantitative assessment of outcome. The rationale underlying QALY assessment is explicit that the value of a QALY is the same no matter who it accrues to: there is no moral component in the calculation. And nor is there the requirement that efficiency of QALY allocation be the sole (or even most important) driver of decision making.

A QALY calculation is fundamentally contingent on the interaction of the intervention with the people being intervened on. Someone’s capacity to benefit (which is what QALYs measure) depends not just on their characteristics but also those of the intervention. Absolute ranking of QALYs as an empirical assessment of someone’s ‘value’ based on their health is therefore meaningless: a different intervention on the same set of people can result in a totally different estimate of outcome.

Consider if rather than five people on the crossing there was only one (and still two siding workers). A pure utilitarian consequentialist would switch from ploughing the train into the siding to smashing it into the crossing. But this doesn’t mean she has suddenly changed her mind about the value of the lives of the people involved, merely that the situation and therefore the most efficient outcome, has changed.

QALYs don’t ascribe a value to someone’s life. They are accounting tokens, providing a (perhaps flawed) quantitative estimate of health outcome in a specific circumstance – usually that of evaluation of an intervention relative to an alternative in an identifiable group of people. This is not to say that some people might not be harmed by a decision based on a QALY assessment, but that, of itself, does not make the decision unfair or unjust.

Alongside utilitarian efficiency and QALYs, egalitarian considerations of fairness and equity, distributional factors, affordability, and political priorities may (and often do) feed into the decisions that are ultimately made.