NEVER MIND THE WIDTH, FEEL THE QALYS
Leibniz... cherished throughout his life the hope of discovering a kind of generalised mathematics, which he called Characteristica Universalis, by means of which thinking could be replaced by calculation. “If we had it,” he says, “we should be able to reason in metaphysics and morals in much the same way as in geometry and analysis.” “If controversies were to arise, there would be no more need of disputation between two philosophers than between two accountants. For it would suffice to take their pencils in their hands, to sit down to their slates, and to say to each other : Let us calculate.” (Russell, 1961) __
Increasing pressure on limited medical resources leads us to expect inevitable rationing of medical services. It is often suggested that
it is better to ration services... explicitly rather than implicitly, and by rational means rather than on an ad hoc basis.
One means of attaining the rational ideal is through seeking to maximise the number of Quality Adjusted Life Years (QALYS) to be gained from any given expenditure.
As set out in a recent article by Singer, McKie, Kuhse and Richardson the QALY process works by allocating a quality-of-life weighting to each lifestate; a ‘normal’ life counts as 1, and any medical condition (or any disability, which is here treated as the equivalent of an incurable medical condition) results in a reduced score. The number of QALYs gained by each intervention is then calculated, using the formula
QALY gain for intervention =_((final QALY weighting) * years of life with intervention) - ((initial QALY weighting) * years of life without intervention)__
The prospective gains in QALYs to be expected from competing interventions are then compared and ranked. One example given is "Karen", a wheelchair-bound paraplegic with severe pain, who has a quality-of-life score of 0.5. In competition for a person without any disability for a lifesaving treatment one would then multiply the expected years of life gained in each case (40, say) by 0.5 in Karen's case and by unity in the case of the person without disability, giving 20 QALYs for Karen and 40 for her hale competitor. Considered logically, the article argues, the greatest benefit would thus in Karen’s case come from giving the treatment to the nondisabled person.
The effect of the QALY loading is thus that people like Karen who have been disadvantaged once by their disability will be statistically discounted in resource allocation and will in some circumstances be disadvantaged again. This effect is known as "double jeopardy".
Before addressing the question of disability I would like to pose a different thought experiment.
The QALY equations also obviously favour people with longer expected lifespans; indeed, Singer et al regard preference for candidates with longer lifespans as obvious, intuitive, and less troubling than preference for candidates without disability - if you have a choice of saving a child or an old man, you obviously choose the child.
If expressed as a principle, however, or an equation, the goal of maximising QALYs has other consequences that are less intuitive. One of the most clearly established features of Australian epidemiology is that Aborigines have a sharply lower life expectancy than other Australians - at birth, an Aborigine male has a life expectancy of fifty-three compared to a general Australian male figure of seventy-three. There is, let us say, a single available life-saving intervention available in the neonatal intensive care ward and there are two possible candidates, Yulipingu or Bruce. If one calculates the QALY yield of saving each life, it is clear that giving the treatment to Bruce is a third more efficient than giving it to Yulipingu. If one follows the logic of the QALY system, it is not unethical, or at least not unfair, to favour whites over blacks in the allocation of health care whenever the scales are otherwise equal.
I must make it clear that Singer et al do not use the Yulupingu/Bruce analogy; indeed, Singer has clearly stated elsewhere that he regards race-based discrimination against Aborigines as unacceptable. I cannot see, however, anything in their system of calculation that would avoid this consequence. If one favours decisions based on lifeyears, and if shorter lifespans correlate, as they do, with poverty and discrimination, then a colourblind test will favour whites over blacks now and for some years to come (we can certainly expect the black-white gap to be reduced over the next fifty-five years, but as Aboriginal children are being born now into conditions of disadvantage that will continue to affect their lifespan even fifty-five years into the future we cannot rationally predict with any degree of certainty that it will be eliminated or reversed). We can believe, as I hope we all would, that we should all work to remove the differential; the ethical question, nonetheless, is whether one can equitably take account of that differential while it continues to exist.
It is true that Singer et al are not necessarily recommending that the logic of QALYs be followed undeviatingly or in all cases; the authors caution that
There is more to overall utility than health-related QALYs, and it is plausible to suppose that tilting the balance of healthcare towards the more disadvantaged members of society will reinforce feelings of concern and sympathy and lead to a more compassionate society.
They do say, however, that following that logic would not be unfair.
...we could understand a preference for avoiding double jeopardy, although double jeopardy is not in itself unfair.
I believe, after some reflection, that if the logic of QALYs dictates that, other things being equal, people without disabilities should be assisted above people with disabilities and whites should be assisted above Aboriginals then these outcomes demonstrate that QALYs are not a fair method of allocating resources and should stimulate a closer examination of the method's underlying premises.
Singer et al do not, as I say, appear to contemplate QALYs being adjusted for anything other than disability or life expectancy. This would appear to amount to prejudice against disability, as a category. They do not propose to adjust QALYs for quality of life in any other respect; QALYs are, as far as the discussion goes, to be blind to temperament, income, sex, gender, weight, looks, habits, and legal status, at least insofar as these do not affect life expectancy. Any actual or assumed measurement that measures quality of life will be sensitive to these factors; any measurement that excludes them is incorporating assumptions from outside the model and cannot claim to be simply translating community wishes into a resources distribution formula.
It is certainly possible to insert an extra rule that says the calculation will depart from the model equation and work on the basis of a standard life expectancy where the actual life expectancy is shorter than the standard when this is due to race or poverty but not when it is due to age or disability. The question must then be why only some categories of disadvantage are seen as relevant, or as natural, or as inherent.
There is no inevitability about this restriction to ‘medical’ conditions. It is feasible, though not necessarily desirable, to use a wider range of variables. One Oklahoma neonatal facility, for example, used a formula (determining which babies with spina bifida would be treated) that combined physical and social factors - QL=NE*(H+S), where QL is quality of life, NE is natural endowments, H is the contribution from home and S the contribution from society, a formulation that meant
the treatment for babies with identical ‘selection criteria; could be quite different, depending on the contribution from home and society.
and which meant, more specifically, that children whose parents had a less than desirable
family income, family debt, employment and employability...[and] educational level
were not recommended for treatment.
Whatever the practical effect of the Oklahoma formula, it was in some ways more flexible than the Singer et al scheme in that it did provide some scope to alter the rating through social changes; a higher value for S, the societal contribution, would balance either a lower value for NE or a lower value for H. In the Singer et al model disability itself is seen as discrete, quantifiable, primary, and inherent, unaffected by circumstance. The theoretical developments in the field of disablement over the last twenty-five years, in which disablement is seen not as a quality inherent in the person but a statement of the relation between the person and his or her environment, are ignored. The quality of life of a person in a wheelchair is affected, among other things, by the accommodations society makes to meet the needs of such people - by pension levels, curb cuts, bus lifts, and societal prejudice. Singer et al regard all these as at best fixed and unalterable and at worst irrelevant. The authors would, I believe, differentiate the handicap caused by racism, as something that 'we can and should try to alter', from such things as incurable confinement to a wheelchair on the grounds that the latter is ex hypothesi unalterable; I would regard that as in some respects a weak distinction, in that in both cases the physical factors are unalterable but the experience - the lived perception of being an Aboriginal or wheelchairbound - can be improved through, among other means, both the removal of prejudice and the expenditure of money.
If disability was seen as a variable rather than a fixed point this would complicate, even in theory, the generation of QALYs. The method suggested relies on asking people
how long a period of life in the given health condition they would be prepared to trade for one year of normal health.
If a range of amenity levels need to be accommodated for each 'given health condition', depending on one’s view of the likely development of services for people with disabilities, this requires a more difficult equation. What, similarly, is 'normal health'?. The concept refers more to a platonic entity than to the actual health norm, under which in Australia illness is almost universal and certainly normal. In the 1989-90 National Health Survey, 75% of Australians reported having taken health-related actions (ranging from medication to hospitalisation) in the previous fortnight, 70% reported having experienced one or more illnesses or injuries, and 66% reported that they suffered from long-term conditions. Over fifteen per cent (15.6%) regarded themselves as having a disability and 11.5% regarded themselves as having a handicap. Nonetheless, 79% of those surveyed perceived their health to be good and only 4.5% thought they had poor health (and negative responses were not necessarily correlated with disability); that is, most people who describe themselves as disabled or handicapped think of themselves as reasonably healthy, and most people who describe themselves as reasonably healthy have a continuing condition affecting their quality of life. When people are asked to compare their concept of disability with their concept of normal health, they are likely to be comparing a media-generated horror with a media-generated fantasy.
Another difficulty is more fundamental. In the example given,
.... Michelle says - and means it - that she would gladly give up ten years of life as a quadriplegic for a single year in normal health. This makes her QALY score 0.1.
Can any real meaning can be attached to the words "and means it" in connection with choices that are by definition impossible of achievement? I cannot see that Michelle's comparison differs in any meaningful way from my stating that I would gladly give up a year of life as a health bureaucrat for one crowded hour of glorious life as, let us say, Madonna. If, within the range of the possible, a representative population was asked to rate its quality of life against what that quality would be if they became richer or poorer, would the ratings be invariant in each case? If not, why should these ratings not also be included in QALY calculations? It would certainly make sense in public health terms, not simply in QALY terms; rich people have longer life expectancies, and giving poor people more money is thus a perfectly defensible medical intervention.
It is also not clear what function the word "makes" has in the quoted sentence. In context, it suggests consent to a negative life loading. If this consent existed, there would be neither a moral dilemma nor a practical problem; Michelle would be knowingly agreeing to stand aside in favour of the nondisabled person. The alternative possibility is that a decision taken in the context of one theoretical comparison is taken also to bind Michelle in later more practical dealings. A fable by La Fontaine covers this issue precisely. A poor woodcutter is lamenting his lot, proclaiming loudly he would be better off dead. Death hears him, steps up, and makes the offer. The woodcutter drops his axe and runs for his life. The moral is “Plutôt souffrir que mourir, c’est la devise des hommes’ - Rather suffer than die, that’s the motto of humanity. We do voice different, and inconsistent, preferences in different circumstances.
More precisely still, Ms Carrie Coons expressed to friends and relatives her unwillingness to live on as a ‘vegetable’. At the age of 86 she was hospitalised after a massive stroke. Initially able to speak single words, her condition deteriorated until she was completely unresponsive. Her daughter applied to the court for permission to terminate care. The court noted that PVS was an irreversible condition, noted that PVS patients were incapable of consciously experiencing or appreciating life, and noted also that the prevailing medical view, as evidenced by the positions of professional societies, supported the withdrawal of artificial feeding from such patients. It was ordered that nutrition be withdrawn. Before the decision could be implemented, however, Ms Coons woke up. She started eating and speaking. She retracted her previous theoretical commitment to termination. The court 'ordered further neurological and psychiatric assessments'.
It is possible to hold the opinion - an ethicist has voiced it - that Carrie Coons' second thoughts on euthanasia were not objectively rational, and need not be heeded;
Her capacity for assessing evidence has been affected by her own bizarre experience...Courts should resist the temptation to oversimplify this job by treating a patient's utterances during brief periods of consciousness as determinative.
It is still necessary to note, however, that her utterances were not consistent, and it is possible to observe that this is hardly surprising.
In the example given, Michelle determines her own weighting. As Singer et al are aware, QALYs are not normally generated by asking people with disabilities the value they place on their own lives but rather by asking people in "normal" health how they would compare their present state with hypothetical states of ill-health. Even within the present article it is not suggested that Karen's 0.5 loading is her own choice. This is not an inevitable feature of QALY use, and is possible to evade the issue by hypothesising a non-specific means of determining whose preference shall count for what. It is certainly not suggested, however, that the hypothetical lifesaving operation will be offered to the person who at the time bids highest on their QALY score.
All these considerations threaten to introduce complications that may make it more difficult to encapsulate all necessary information into a single figure. Singer et al avoid many of these disputes over the practical difficulties of drawing up QALY ratings by hypothesising, after the manner of ethicists and economists, a perfect measure.
...let us assume that we have a suitable measure of the value that continued life holds for the person whose life is at stake. This value may be identical to what is now measured by QALYs, or it may be a measure of strength of preference, or it may be something else altogether. For simplicity, we shall refer to it as a person's 'interest in continued life'.
Even this unspecified perfect measure, however, cannot avoid all possible objections. Any conceivable valid measure must still assume that the value, or preference, or interest, is not only measurable but scalar, inherent, and commensurable - scalar, in that to allow the imposition of arithmetical procedures the difference between a 0.6 QALY and a 0.4 QALY must be the same as that between 0.4 and 0.2, or 0.6 and 0.8; inherent, in that in order to be stable the interest must be in undifferentiated 'natural' life rather than in life under possible alternative social circumstances; commensurable, in that any two lives of similar ratings (and any life and any hundred, or thousand, or million headaches) must be interchangeable. None of these assumptions are either probable or particularly helpful.
It is possible, of course, to use QALYs in more limited ways that do not necessarily discriminate against people with disability. It is possible to imagine a management tool where different interventions are compared in their effects on a standard life rather than the same intervention being compared in its impact on different lives. They might be used to compare, for example, two interventions, one that resulted in a longer lifespan but more pain or more impairment and one that resulted in a shorter lifespan without either. Even here, however, it is not necessarily self-evident that a QALY approach would be helpful.
The appeal of the QALY is that it promises to replace a system where health funding priorities are set by historical accident, vested interests, prejudice, and inertia by a 'rational and explicit resource allocation process'. It is proposed that where politics, the market, and social norms now compete and combine to place values on medical outcomes this should be replaced by rational calculations based on agreed principles. It is still not clear that this would be a good thing.
“The methods of science,” he [B. F. Skinner] wrote, “have been enormously successful wherever they have been tried. Let us then apply them to human affairs.” Let us, indeed. Few of his readers observed that of the two parts of this claim the first is false and the second a non sequitur.
The success of the exact sciences over the past three hundred years have led us to accept unquestioningly the matter-of-fact application of the operations of mathematics to events in the real world, and may have concealed from us the conditions that have to be fulfilled before these concordances apply. When dealing with gravity, or heat, or weight it is possible to assume that behind the accidents of colour, shape, cost, gender and position there are universal regularities, and that these regularities can be simulated by mathematical formulae. Two units of heat added to a further two units of heat are in most respects commensurable. With socially determined indices, however, there is no higher reality behind the veil of appearance. A preference, or an interest, expressed in a particular case is not an instance of an underlying rule. It can be complete in itself, and need have no linkage, let alone a mathematical correspondence, with any other preference or interest. In the realm of numbers, if A>B, and B>C, then A>C; a voter can, however, prefer candidate A to candidate B, candidate B to candidate C, and candidate C to candidate A. Any requirement of consistency between preferences must be imposed from within the discipline of ethics, itself socially constructed; it cannot conceivably be extracted from the raw data.
It would thus follow that Michelle's preference for a single year in normal health rather than ten years of life as a quadriplegic cannot be taken to determine her position in the different choice between having or relinquishing a lifesaving intervention.
If there are no underlying principles, furthermore, then we must ask why the determination of preferences through survey should be preferred to the more complex, if less consistent, processes of politics and the market. Each of these has its own means of processing information about preferences, and if they pretend to less precision they have at least the virtue of being able to register failure. Politicians can lose elections and businesses can go broke, while for economists and ethicists any variance between prediction and outcome is accounted for by their having been let down by the inexplicable unwillingness of the population to adopt their model in its entireity. Political processes, moral guidelines, and social habits represent, among other things, alternative modes of collapsing information, and it is exactly their acceptance of inconsistency that makes them more capable than the scientific ethicist of dealing with the level of complexity involved in the achievement of a “compassionate society”.
If there are no underlying principles, a statistical and QALY-based approach does not represent the elimination of irrelevant information to allow the underlying regularities to emerge more clearly; it represents the elimination of the only information there is or can be, and a destruction of material that should correctly be taken into account in decision-making. QALYs do not represent a distillation of the raw data, still less an enrichment. They represent an impoverishment. This can be readily seen for the individual; if I am offered a choice of two 20-QALY treatments, I can reasonably wish to undo the mathematical transformation in order to know which of two treatments with identical QALYs has fewer years but less pain. A policymaker can, and should, ask much the same thing. The QALYs number cannot hold the information that it is necessary it should if the maximum satisfaction of preferences is to be our goal.
We must give up the dream, stretching back to Leibniz, of being able to lift our disputes out of the realm of argument into the realm of logical, arithmetical, demonstrable, unarguable, eternal truth. If our health preferences are irreducibly socially based and socially expressed then the remedy for deficiencies in present health arrangements must be to have better and more inclusive processes for involving social groups, social forces, and social interests in the allocation processes.
There is, of course, no guarantee that these processes will produce either a logical or a ‘morally right’ answer, or even that they will produce a more favourable result for people with disabilities than a QALY-based system. Present political and social arrangements, after all, produce a clear tendency for white middle-class middle-aged non-disabled voters to favour their own interests, and this tendency may be successfully mobilised against would-be reformers. It is not clear, however, exactly how these undesirable existing tendencies are to be suppressed simply through the use of QALYs. If I say that the remedy for a problem based in maldistribution of power is to adjust the distribution of power, Singer et al may very well object that this remedy is impossible of achievement, or at least unlikely. This may very well be so. Such an objection also implies, however, that while white middle-class males will not relinquish power in a political contest they will nonetheless surrender the fruits of that power without demur if the mathematics of the situation tells them they should - that they will listen to rational argument even when it goes against their own interests, and that even when the veil of ignorance is withdrawn they will agree to an apolitical asocial QALY system that disadvantages them.
If QALYs can elevate the tone of the discussion to that extent, the supplementary question arises of why they are to be confined to the area of health - why decisions about roadwidening, or taxation levels, or industrial policy, or censorship, or any of the items that we at the moment rely on political interaction to settle, should not fall (if we assume, after the fashion of ethicists, that we have a suitable measure of the value that these decisions add or subtract to quality of life) into the realm of mathematics. This is not a frivolous objection; there have always been thinkers who believed that government could and should be handled that way - that there was an independently existing right apolitical answer to such dilemmas, an answer that could be ascertained with certainty by those not blinded by political interest or deflected by political influence, and that once this answer was pointed out it would win assent because it would be scientifically demonstrable. I cannot see that this vision of the future is very much less utopian than one resting on changes in power relationships; rather, I see it as the utopia of H.G. Well's samurais - government by the advanced minds, experts and scientists - and as such only sustainable if resting at some remove on an authoritarian and undemocratic base.
I do not argue for the total abandonment of rationality in favour of political action, or even against the use of mathematics in health evaluation. Mathematics is of greatest use, however, in areas where its complexity is comparatively trivial - comparing a procedure, say, that produces three deaths in ten with another that produces one death in ten - and least defensible in the areas of macroallocation where it would be most useful. It is simply not legitimate to assume without further argument that case-by-case adhocery driven by political processes and moral or social beliefs is intrinsically inferior to rule-governed analysis.
Rather, any attempt to compress ethical, financial and social data on to a single index is intrinsically absurd. The differing imperatives of each field do not cohere, the individual data elements that could support such a measurement are too protean to be fixed by a number, and the amount of information that is relevant even at fairly basic levels of comparison overloads unitary measures and defies valid mathematical simplification.
Russell, B., History of Western Philosophy, Unwin, London