The price of life? The good, the bad, and the ugly of valuing health
There is a certain kind of person I meet a lot at conferences. They’re often patient representatives - engaged, educated people who have dedicated enormous amounts of time to improving medicine for the people they represent. They often recognise the importance of health economics, but as a general rule they have one big issue with it: The QALY, and more particularly, the EQ-5D.
QALY stands for quality-adjusted life year, and it’s the main health outcome used in health economics, at least in the UK. As you’d expect, it’s a measure that combines length of life and quality, and mathematically it’s very straightforward. You multiply the length of life, say 5 years, by a quality adjustment, also called a health utility, that ranges between 0 (death - although scores can go negative, i.e., worse than death) and 1 (perfect health). So five years of some condition with a quality of life of 0.8 is equal to 4 QALYs (0.8 x 5).
The nice thing about QALYs is that quality of life is valued in a way that trades off length against quality. To do this, rather than simply ranking health conditions, you ask people how much time living with one condition they would be willing to give up to avoid a worse one. This means that 4 QALYs always represents that same amount, irrespective of whether that is 4 years of perfect health or 8 years at a quality value of 0.5.
Where it becomes problematic, though, is in trying to assign that quality value to a named condition, for example cancer. The values themselves are developed using something called “health states”. These are lists of health problems of varying severity, for example: “you can do usual activities, experience moderate pain, take care of yourself with no problems, have extremely impaired mobility and are moderately depressed”.
By getting patients with the health condition of interest to complete a questionnaire that produces a result like the one above, we can convert a health condition with a name into a generic health state and assign it a quality of life value.
It’s a system that works quite well for the purposes of comparison, but has some pretty obvious flaws. The health state I gave as an example above is taken from the EQ-5D, which is the most commonly used questionnaire, and the one that must be used in NICE health technology appraisals. Every variant of that example (with no, some or extreme problems in each category) gives a maximum 243 possible health states. That number is sufficient to cover the essential differences between health conditions pretty well for the purposes of decision making. That said, it is not hard to see how a five category questionnaire can look pretty inadequate when you’re a patient.
So that’s why the EQ-5D, and QALYs in general, aren’t a patient representative’s favourite thing. They don’t capture the full experience of living with a condition. At the moment the general case for them is that they aren’t designed to. That despite being called “quality of life” values, they are only designed to capture what is called health-related quality.
In other words, if a health condition includes some element of health that isn’t covered by the EQ-5D, that’s only a problem if the health condition would be valued much better or worse, relative to other conditions, if that element were included. In that case, the decisions around making treatments available might change because a new medicine that improves the missing element might not look cost-effective when it actually is, which risks denying patients effective care and reducing how effective the health system is as a whole.
And it is easy to identify parts of a health condition that are not fully covered in the EQ-5D. Anyone with experience of an illness can find something missing. Beside aspects of what is classically thought of as medical problems, there are also emotional, social and economic benefits. Why not include them? These are all important to patients.
This is where it starts to look really messy. If we don’t limit quality of life to just pure medical benefit then there is a risk that we start using health budgets to subsidise other things. For example, treating conditions that get people back into work looks good for the economy, but should we therefore spend more money treating these younger, less sick patients who can plausibly get better than we do on dementia patients, most of whom are are retired anyway?
It is tempting to continue adding extra benefits to the treatment you want to approve. This happens a lot in pregnancy, where the choice of whether to count the health of an unborn child among your benefits is strongly predicted by whether that makes your treatment look more or less cost-effective. Besides raising questions about whose health matters, this approach usually also ignores that every other treatment you are comparing against will have its own non-health benefits. If the threshold we used to measure cost-effectiveness included all those extra benefits it would be higher, and the bar would be raised yet again.
That’s not to say that we shouldn’t include extra benefits, and methods exist to do so in a small number of cases. Doing so tends to bring its own challenges, though, which I’m planning to do another blog post on soon.