Friday, September 26, 2014

Where QALYs Go Wrong

My paper 'Against "Saving Lives": Equal Concern and Differential Impact' defends the use of QALYs (Quality-Adjusted Life Years) in medical resource allocation against several traditional objections. But along the way, I note several respects in which (it seems to me) not all life years -- even in perfect health -- are equal, and hence a straightforward QALY-maximization approach falls short.  I'll briefly outline them below, and invite readers to suggest any further examples I may have missed...
(1) Non-persons and new persons.

Suppose you could either (i) save the life of a ten year old child, (ii) save the life of a ten week old infant, or (iii) conceive and bring into existence an entirely new child. Supposing that the individuals in question will face no future health risks until they die of old age, it appears that the options are listed in increasing order of QALYs thereby "gained". But it seems entirely backward to think that this corresponds to the moral desirability of the options -- (i) is clearly better than (ii), which in turn seems better than (iii).

What we really care about is helping individuals rather than just maximizing QALYs for their own sakes. Bringing additional good lives into existence -- as in option (iii) -- is impersonally good, but arguably is not a benefit for the new person (say if we judge non-existence to be incomparable to good existence). At any rate, it doesn't seem to have the same moral urgency as preventing harms, by which I mean improving the lives of people who will otherwise exist in a worse-off state.

Further, in preferring (i) to (ii), we reveal doubts as to whether future QALYs really benefit an (even already-existing) infant as much as they benefit an older child. (Put another way: it seems that the older child is more harmed by early death than the infant is.) This is arguably due to the relative lack of psychological continuity between the infant and their (possible) future stages. For example, future good experiences in my body only benefit me if I will be the one experiencing them, and the normatively relevant sense of diachronic personal identity here is one that's grounded in psychological continuity: overlapping chains of memories, beliefs, values, intentions, etc. Infants don't have a lot of these, so they are only very weakly connected to their future selves, and so don't lose as much if those future selves are prevented from existing (by early death).

Put this all together, and we get the result that QALY gains to non-persons and to new persons should be significantly discounted.

(2) The Best Years of Our Lives?

Different periods of life are likely to be more or less "central" to the realization of our core life projects and values. For example, perhaps life-years in the period from late adolescence through middle age are generally more important/beneficial for the person living them.

(2a) Positive externalities: Not something I discuss in the paper, but the "central/productive" years of life are also arguably the ones with the greatest positive externalities or "social value" for others. While there's no principled basis for distinguishing direct vs. indirect beneficiaries in general, appeals to social value in health care allocation open up a whole new can of worms, so I won't pursue that further here (but you're welcome to in the comments!)...

(3) "Chunking" effects?

Much of what really matters in life requires significant "chunks" of time to realize. While this can help explain some intuitions about, e.g., favouring a few larger QALY benefits over a larger aggregation of smaller benefits, such intuitions seem to stem from the false assumption that additional QALYs can only help one to complete an important project if the entire project occurs (from start to finish) within the new period of time granted. But of course, even a small life-extension may allow people to complete a significant project that they were already part-way through. So in practice, as is so often the case, I don't think that chunkiness actually ends up making any practical difference.

(4) Closure.

The value of closure -- saying one's "goodbyes", making final plans, etc. -- might mean that we should give extra weight to even fairly short life-extensions for patients who would otherwise die unexpectedly early (i.e. with no prior warning).

Any further thoughts or suggestions?


  1. Interesting stuff!

    With regards to (1), do you have any thoughts on the size of the discount you think should be applied? I wonder how (if) this would impact our perceived value of foreign aid and effective altruism-recommended charities, given that much of the benefits of (e.g) reducing malaria derive from reducing morbidity and mortality of children under 5?

    In terms of (3), I don't quite follow. Are you saying that 'chunking' is *not* an argument against QALY maximisation?

    As for (4), does this imply we should value emergency-room interventions over interventions that provide an equivalent period of life-extension for those with terminal illness?

    1. re: 4, yes, that's a nice example of how that principle might be put into practice.

      re: 3, yeah, I'm suggesting it isn't an argument against QALYs as a practical policy proposal -- it's merely to note another way that QALY-maximization might in principle come apart from benefit-maximization (in a way that's basically unknowable, and so unable to be acted upon).

      re:1, yes I think that is some reason to slightly down-weight AMF and similar charities, at least so far as their life-saving effects on the very young are concerned. (Morbidity still matters just as much, since if an early illness negatively affects one's development, that may lead to significant harms to the future person.)

      Not sure how best to determine the appropriate discount rate. One idea would be to estimate the degree to which very young children (at different ages) seem to exhibit psychological continuity. A rougher estimate could be obtained by determining the relevant end-points (perhaps a newborn shows the barest glimmers of psychological continuity, and a four-year old has sufficient for full personhood), and assume linear development between them for simplicity (so that e.g. a two-year old would receive a 50% discount).

  2. (I posted this once, but it might have been lost. Sorry for any duplication.) Looking forward to reading the paper. For now, I'm curious about the title of this post. Is your view that qalys *themselves* "go wrong", or just that qaly *maximization* goes wrong? That is, do you hold that qalys are a good metric for measuring *health* improvement, but reject that equal units of health improvement are equally morally valuable?

    1. Hi Paul, yeah I guess the latter characterization is better. "Health" seems a rather nebulous concept (e.g.are disabilities properly characterized as a form of ill health?), but -- partly for that reason -- I'm also inclined to doubt that it's a particularly useful or important concept. (Do you disagree? I'd welcome arguments to the contrary.)

      Really I'm just focusing on the core moral issue of improving wellbeing, and noting some ways that QALYs (whether we call them "health improvements" or not) fail to track morally relevant benefits.

    2. Hi Richard: I think the "ill" in "ill health" might add an unwanted connotation, but yes I'm inclined to say that (at least most) disabilities are properly characterized as a form of diminished health (i.e. a health decrement, in Chris Murray's favored terminology).

      I do agree with Broome and others that qalys don't "measuring health", they measure the value of health. But I haven't made up my mind on whether health's value turns on its impact on well-being, or on something else. Broome says the former; Hausman says the latter.

      Do you know Broome's paper "Good, Fairness, QALYs"? Hits on some of these same themes, including responding to Harris. I've uploaded it here, in case you don't have it yet:

      I also think it's useful to follow that paper with pages 261-2 of *Weighing Lives*.

      Again, very much looking forward to reading your paper! I was not myself much convinced by Bognar & Kerstein (2010).

    3. Ah, that looks helpful, thanks for the link!


Visitors: check my comments policy first.
Non-Blogger users: If the comment form isn't working for you, email me your comment and I can post it on your behalf. (If your comment is too long, first try breaking it into two parts.)