The paper proposes a new method of researching public opinion for the purposes of valuing the outcomes of healthcare interventions. The issue I address is that, under the quality-adjusted life-year system, disabled patients face a higher cost-effectiveness hurdle than able-bodied patients. This seems inequitable. The author considers the alternative approaches to valuing healthcare interventions that have been proposed, and shows that all of them face the same problem. It is proposed that to value an outcome, instead of researching the general public, the population that is to be targeted with the intervention should be researched.
- Allocation of healthcare resources
- health economics
- health technology assessment
- healthcare economics
- healthcare resource allocation
- philosophical ethics
- quality of life
- quality/value of life/personhood
Statistics from Altmetric.com
- Allocation of healthcare resources
- health economics
- health technology assessment
- healthcare economics
- healthcare resource allocation
- philosophical ethics
- quality of life
- quality/value of life/personhood
When we research how the public values various health states, we find that non-functionally impaired states are preferred to functionally impaired states. Jonathan Wolff quotes a UK study of more than 1000 people who were asked how they would regard being ‘confined to a wheelchair for the rest of your life’. Approximately 28% of people said that it would be as bad as death, and a further 24% regarded it as worse than death. In other words, more than half of the respondents thought being in a wheelchair was as bad as death or worse.1
This is important because healthcare funders use public opinion research to guide their decisions about whether to fund treatments that produce such health states. ‘Health technology assessments’ are catching on around the world. One of the earliest adopters was the UK, where the responsible body is the National Institute for Health and Clinical Excellence (NICE). NICE assesses an intervention according to how many quality-adjusted life-years (QALYs) it produces. QALYs consist of incremental life years adjusted for quality of life.
I am concerned about how this affects people with disabilities. Suppose you have got two groups of patients of average life expectancy, the only difference being that one group is disabled and the other not. For simplicity, suppose that both groups have life-threatening conditions such that life expectancy without treatment is 1 week, but the prognosis for both if they are given life-saving operations is 40 years, during which they will have their preoperative quality of life. Suppose there is no alternative treatment for either group. In view of the research reported above, let us also suppose that the quality of life of a disabled patient is deemed to be just 15%, whereas that of the non-disabled patient is 100%. This means that even if the life-saving operations produce the same number of incremental life years for both groups of patients, they will only produce six QALYs for the patients with disabilities compared with 40 QALYs for the non-disabled patients. So we gain more QALYs by saving the lives of the non-disabled patients than by saving the lives of the patients with disabilities.
NICE then calculates the cost per incremental QALY of the treatment, and funds treatments whose incremental cost per QALY falls below a given threshold. However, because the intervention for patients with disabilities produces fewer QALYs, the cost per QALYs will be higher even if the operation costs the same and produces the same increase in life expectancy. So it could be that the operation for non-disabled patients gets funded whereas the operation for disabled patients does not.
I suggest that a disabled patient who lost out by this kind of comparison could justly complain of unfair discrimination. John Harris calls it a problem of ‘double jeopardy’: ‘QALYs dictate that because an individual is unfortunate, because she has once become a victim of disaster, we are required to visit upon her a second and perhaps graver misfortune. The first disaster leaves her with a poor quality of life and QALYs then require that in virtue of this she be ruled out as a candidate for life-saving treatment, or at best, that she be given little or no chance of benefiting from what little amelioration her condition admits of.’2
Healthcare commissioners will no doubt respond that if the conclusions of a QALY-based analysis look discriminatory, their committees have discretion to set the QALY-based analysis aside and make a different decision.3 However, the exercise of such discretion would be an admission that standard QALY theory is inadequate. If we can come up with a better analysis to feed into committee decisions, we will reduce the chance that a misleading analysis could lead a committee into a wrong decision. Also, we expect public bodies to be transparent and consistent. The tighter the fit between their decisions and the supporting analysis, the more transparent and consistent those decisions are likely to be. In contrast, if a decision diverges from the conclusions of the supporting analysis because decision-makers are influenced by ethical qualms that are not reflected in the analysis, the resulting decision will be more difficult to explain, more difficult to understand and vulnerable to accusations of inconsistency.
In what follows, I consider the alternative approaches to valuing healthcare interventions that have been proposed, and show that all of them face the same problem. Then I propose a new way of researching public opinion, which mitigates the implication that disabled states always tally with lower quality of life.
Here is one reason to doubt that the quality of life associated with disability is as poor as traditional public opinion research suggests. If quality of life is measured in terms of emotional states such as contentment, the quality of life of people with disabilities often matches that of non-disabled people. They are supported by surveys that show that following an accident or illness that causes a disability, within a few months people's subjective contentment often reverts to what it was before.
For example, Lulé et al4 found that quality of life in patients with locked-in syndrome is often in the same range as in age-matched healthy individuals. Diener et al5 mention that according to Feinman,6 people who were blind did not differ in happiness from those who were able to see. Also Silver7 found that individuals with spinal cord injuries reported strong negative emotions 1 week after their crippling accident, but 2 months later, happiness was their strongest emotion.
However, against this, NICE will point to its own surveys, showing that despite the above fact, the public places a low value on functionally impaired states. Furthermore, such surveys are likely to produce similar results when you ask people with disabilities. In other words, although the emotional state of people with disabilities matches what their emotional state would be without their disability, people with disabilities are not as satisfied with their state as non-disabled people are with theirs. For example, Schulz and Decker8 report that 100 paraplegic/tetraplegic respondents gave a mean score for life satisfaction of 10.76 on a scale of 0 to 18, which is lower than the mean scores for other groups ranging from 12.2 to 13.35.
In light of their survey results, how can healthcare funders legitimise a change in policy?
One seemingly obvious alternative is for funders to switch from researching public preferences for a state and instead research patients' subjective states of contentment in a given health state. A proponent of this option is Daniel Kahneman, who has proposed that health states should be valued according to experienced utility,9 10 but there are at least two issues with this.
First, it seems to me to be inconsistent with a more general point about our rights to healthcare. In principle it seems that I should have the same entitlement to healthcare as the next patient, regardless of my projected state of contentment following treatment (other things being equal). For example, if I am temperamentally grumpy or neutral in my moods, that does not mean I am less entitled to treatment than someone who is temperamentally more upbeat. So my subjective states do not influence my entitlement to treatment one iota. The same applies to groups of patients. Second, if the value of a health state is determined by the subjective contentment of those in that state, this would not explain why it is right to fund preventive programmes to reduce the prevalence of disabilities, nor why it is right to remove a disability if we can.
A second alternative to the current system that has been canvassed is the capabilities approach, which says that healthcare policy should aim at maximising a range of basic abilities. For example, Paul Anand11 argues that the capabilities approach provides a rationale for some of our existing policies of healthcare resource allocation that would be difficult to justify from other perspectives. Cookson12 has proposed that QALYs can be reinterpreted as an index of the value of an individual's capability set. Joanna Coast and colleagues offer a capabilities-based self-report measure of wellbeing for older adults.13–15
If we are to see this approach as distinct from the ones we have considered, we have to emphasise that the list of capabilities may be constructed to an extent without reference to the preferences of the target population. For example, we form some supposedly culture-neutral conception of human flourishing and then we consider what capabilities are required for humans to flourish that way.
However, the capabilities approach seems even less able to support the nuanced approach we are looking for. As soon as you insist that the most valuable health states are those in which a fixed list of capabilities are fully realised, you are forced into a fixed view of disability as a lower value state. That is not surprising given the origins of the capability approach. Amartya Sen16 had seen how people in poverty tend to adapt their preferences and their subjective contentment to their condition, and he wanted an approach to economics that bypassed that adaptive response (which is why the capabilities approach sometimes shuns preference-based utilities).17 Perhaps we are looking at a similar adaptive response in the case of some people with disabilities. However, I do not think Sen would have wanted disabled people or the poor to be at the back of the queue for life-saving operations, so I think in principle he would have supported this search for alternatives.
Constraints on an adequate solution
In fact, all the approaches we have considered have a problem accounting for three seemingly contradictory intuitions regarding possible interventions.
The prevention intuition
It is permissible (perhaps sometimes obligatory) for governments to spend money on public health programmes and safety campaigns to reduce the incidence of disability; for example, osteoporosis prevention programmes, or campaigns to reduce the incidence of deafness in noisy workplaces.
The removing disability intuition
It is permissible (perhaps sometimes obligatory) for governments to spend money curing disabilities.
The life-saving intuition
The disabled are as entitled to life-saving operations as the non-disabled.
Now if you analyse these possible interventions from the point of view of standard QALY theory, or any theory based purely on assigning a fixed value to health outcomes, the intuitions are contradictory. If it is right to spend money on preventing disability and curing disability, according to outcome-based theories that can only be because the quality of life of the disabled is lower than that of the non-disabled. However, if disabled patients have equal entitlement to have their lives saved, given that interventions are valued according to the value of their outcomes, that can only be because the quality of life of disabled patients is equal to that of the non-disabled. The reason is that, as we have seen, if the quality of life of patients with disabilities is judged to be lower, there would be circumstances in which a life-saving operation for non-disabled patients would be funded but a life-saving operation for patients with disabilities would not be (and no situations in which the reverse is true).
In summary, assuming interventions are valued according to their outcomes, these intuitions depend on two different ways of valuing the disabled state. According to the prevention intuition and the removing disability intuition, being disabled is a lower value state than being non-disabled, but according to the life-saving intuition, being disabled has approximately the same value as being non-disabled. Thus the intuitions thus look inconsistent.
In view of this difficulty associated with valuing outcomes, I will note a third alternative to the current system that has been canvassed. This is to drop the assumption that interventions should be valued according to their outcomes. John Harris argues that all patients needing treatment are equally entitled to it, so the healthcare system should give them an equal chance of getting it, for example, by drawing names out of a hat. However, this has counterintuitive consequences. If we have got a choice between an intervention that produces 30 years of total paralysis plus extreme pain for one set of previously healthy patients versus another intervention that produces 30 years of unimpaired pain-free health for another set of patients, all else being equal, most people will feel that the second group should be preferred. Something similar applies if we have got a choice between adding a few days of life for one set of patients versus adding 40 years for another set of patients. Given there is not enough money to fund all treatments, it seems at least some of the time we should select the interventions we fund according to their outcomes.
So I would like to see if we can continue valuing interventions according to their outcomes while at the same time respecting the above three intuitions. This requires us to ask how can we say that being disabled is a lower value state than being non-disabled—yet at the same time, also say that being disabled has approximately the same value as being non-disabled?
The answer I propose is to take a relativised view of the value of disabled life. From the point of view of the non-disabled, disability is a low value state, but from the point of view of someone with a disability, disabled life can be OK. Not as good as being non-disabled, but close enough for many practical purposes—including healthcare resource allocation.
This looks like it is pointing us back to the account based on subjective contentment, but that remains unacceptable for the reasons given above. Instead I would like to canvass the idea that the relevant value of a prospective state can vary according to the perspective from which it is evaluated. In particular, the value of a disabled state depends on whether it is viewed from the perspective of someone disabled or from the perspective of someone non-disabled. Given the findings regarding the subjective contentment of people with disabilities, I would expect to find that their valuation of their disabled state, as revealed in preference research, would be higher than the general public's valuation of the same state.
In detail, my patient preference theory says that the value of the state produced by a healthcare intervention is determined by the target patients' current valuation of that state (not the general public's valuation of it and not the valuation of people in the state). The same applies to the state patients will be in without the intervention—for example, if patients will die or deteriorate without treatment, the value of that possible future state is determined by the patient's current preferences. To value the outcome of an intervention we compare patients' current valuation of the state they would be in following the intervention with their current valuation of the state they would be in without it. These valuations are benchmarked against patients' valuation of their ideal state, generally full health, to produce a quality of life figure. We then multiply up those two quality of life figures by the respective life expectancies, and compare the two figures to derive the incremental QALYs produced by the treatment.
I contend that patient preference theory will match our ordinary pretheoretical intuitions better than standard QALY theory in any case when patient attitudes diverge from public attitudes. This includes the three intuitions mentioned above. In addition, there will be no case in which patient preference theory is worse than standard QALY theory at matching our ordinary intuitions. For these reasons, patient preference theory is to be preferred over standard QALY theory.
First, patient preference theory is as good as standard QALY theory at matching our intuitions regarding prevention and cure. For example, consider a programme to prevent deafness, such as a health and safety initiative in noisy workplaces. The targets of this programme are people who can currently hear. Without the programme, some hearing people will become deaf. Patient preference theory says these individuals' quality of life in that potential state of deafness is determined by the target population's preferences for that state before they enter it. The targets of the programme will prefer a future state in which they can hear to a future state in which they are deaf. So when we calculate a quality of life figure for those who would lose their hearing without the preventive programme, we find their quality of life after becoming deaf is lower than the quality of life they would have had with the programme. This enables us to place a high value on an effective prevention programme.
Second, assuming that a given set of patients with disabilities place a lower value on their state than on a state of non-disabledness, we can show that an operation to remove their disability is cost effective. When we do preference research with these patients, the quality of life figure we derive for the outcome following the operation to remove their disability will be higher than the quality of life figure for the outcome without the operation. This enables us to place a high value on curing their disability.
Third, patient preference theory is also an improvement on standard QALY theory in relation to the cost effectiveness of a life-saving operation for patients with disabilities. The outcome of the life-saving operation is that the patients continue in their disabled state rather than dying. Preference research with these patients might produce a 70% quality of life figure for the outcome of the operation, based on the fact that the outcome is not quite as good for them as an outcome in which they continue to live non-disabled. However, in the vast majority of cases, this will be enough to avoid discrimination. I consider some figures in more detail below, but generally, according to patient preference theory, other things being equal, if a life-saving operation for non-disabled patients is cost effective, a life-saving operation for patients with mild disabilities will also be cost effective (and if the latter is not cost effective, nor will the former be). In other words we have achieved something close to equality between the two groups of patients.
This contrasts with standard QALY theory, in which preference research might only produce, say, a 15% quality of lifei figure for the outcome of the life-saving operation for patients with disabilities, instead of 70%,ii simply because our valuation of the disabled state is based on surveys of the general public rather than people with the disability. As a result, according to standard QALY theory, the life-saving operation for patients with disabilities will be a lot less cost effective than the life-saving operation for non-disabled patients. This is the discrimination we set out to avoid.
It is admittedly counterintuitive that there is potential for a kind of reverse discrimination between patients with disabilities and non-disabled patients. For example, the theory could recommend that we refuse to fund a life-saving operation for non-disabled patients whose treatment will cause a disability, while recommending that we fund a life-saving operation for patients who already have that same disability. In other words, the outcome of the intervention is the same (life with a certain disability), but the policy could be different depending on whether that disability was present before treatment. However, we can justify the discrimination by pointing to the patient preference research showing that the value of life with a disability for able-bodied patients is lower than the value of the same kind of life to patients with a previous disability. Of course, able-bodied patients will not like it any more than anyone else who is refused treatment—but the question is whether they could be brought to see the decision as fair. I suggest it would be easier to persuade able-bodied patients of the fairness of this decision than to persuade disabled patients of the fairness of the discrimination they are vulnerable to under standard QALY theory.
So, the general point is that, for all conditions in which public preferences diverge from patient preferences, patient preference theory will be more likely to recommend a course of action that avoids intuitively unfair discrimination. Standard QALY theory is hostage to public opinion. The greater the divergence between public opinion and patient opinion regarding a disabled state, the more often standard QALY theory will produce counterintuitive results.
A problem with the proposal
However, I have to confess patient preference theory still does not make the life-saving operation for patients with disabilities quite as easy to fund as a life-saving operation for the non-disabled. The issue arises because our patients with disabilities place a lower value on the disabled outcome than on a non-disabled outcome. So patient preference theory still says the life-saving operation for these patients is less cost effective than the life-saving operation for non-disabled patients.
For example, take a cancer treatment costing £16 000. For simplicity, I will assume the cost-effectiveness threshold of NICE is £30 000 per QALY (the argument could be restated without loss of validity for NICE's more complex parameters). To take an example case, if incremental life expectancy produced by treatment is estimated at 7 months for all patients, then patient preference theory will say that the treatment for the non-disabled patient is cost effective while the treatment for the disabled patient is not. The non-disabled patient will get 0.58 QALYs, implying a cost per QALY of £27 000. However, patient research produces a quality of life figure for disabled patients of 70%, then 7 months extra life expectancy will only represent 0.41 QALY for the disabled patients. So their cost per QALY is higher at £39 000. In summary, patient preference theory recommends discriminating between the two groups if the incremental life expectancy produced by our £16 000 treatment falls anywhere between 7 and 9 months. This is an issue.
However, in my defence, first note how precise our assumptions have to be to produce the problematical cases. If we assume treatment adds only 6 months instead of 7 months, then both treatments are rejected, avoiding discrimination; and if we assume the treatment adds 10 months, both treatments are funded, again avoiding discrimination. Something similar applies to cost. If we assume cost is much more, then both treatments are rejected; much less, and both are funded. To get the case in which patient preference theory recommends a counterintuitive act of discrimination, we have to make some quite precise assumptions. So counterintuitive acts of discrimination are very unlikely according to patient preference theory.
Even when patient preference theory recommends discriminating between the two groups, the decision is marginal. For example, for our £16 000 treatment, in the most extreme case when it only adds 7 months of life, patient preference theory says the cost per QALY for non-disabled patients is £27 000, whereas the cost per QALY for disabled patients is £39 000. This is not far over the cost-effectiveness threshold. NICE has been known to approve treatments costing this much. In contrast, in the same circumstances, standard QALY theory says the cost per QALY for the disabled patients is £182 000. This is way over the cost-effectiveness threshold and NICE has never approved a treatment costing that much.iii So in these circumstances, patient preference theory allows a bit of room for manoeuvre to avoid a discriminatory decision, whereas standard QALY theory does not.
Moreover, standard QALY theory has difficulties with a much wider range of cases than patient preference theory. If standard QALY theory judges the quality of life of a group of disabled patients to be 15%, and if our treatment is expected to produce anywhere between 7 months and 3 years 6 months of extra life, standard QALY theory will recommend discriminating between the two groups—non-disabled patients will be funded but not disabled patients. Standard QALY theory only avoids discrimination if our £16 000 treatment is estimated to increase life expectancy by 3 years 7 months or more (all patients get funded), or 6 months or less (no patients get funded). Recall that discrimination under patient preference theory only occurred in cases in which life expectancy was expected to increase by between 7 months and 9 months. This is a much narrower range of discriminatory cases. Therefore, standard QALY theory is likely to recommend discriminating between the two groups of patients much more often.
So patient preference theory is a significant improvement on standard QALY theory. In the vast majority of cases, it recommends better decisions than standard QALY theory (ie, decisions that tally with our intuitions). In all the cases in which patient preference theory recommends a counterintuitive decision, standard QALY theory recommends the same decision. So patient preference theory is not at a disadvantage to standard QALY theory in these counterintuitive cases. What is more, patient preference theory shows these counterintuitive cases to be borderline, allowing some room for decision-makers to exercise their discretion and avoid discrimination, whereas in the same cases standard QALY theory rejects the disabled patient quite definitively without any room for manoeuvre.
For these reasons, anyone who shares the intuition that disabled patients are equally entitled to life-saving operations must prefer patient preference theory over standard QALY theory.
Nevertheless, someone who favours equal treatment for disabled patients might look for another, even more egalitarian, theory that avoids even the small residue of cases in which our theory recommends counterintuitive discrimination. However, I suggest it will be difficult to find any better alternative if we wish to continue basing our decisions on valuing outcomes.
Given this, does patient preference theory not look a bit arbitrary? If we are forced to allow inequality in some cases, does this not mean our pro-equality intuitions are unreliable? Any principle of equality must surely be absolute and overriding. If it turns out that in the case of disabled and non-disabled patients, it cannot be, maybe that is because there is no equality principle in play of the kind our intuitions suggest. So does that not suggest that maybe our pro-equality intuitions are a mistake?
I suggest that the response to this must be to understand the nature of our pro-equality intuitions better. The intuitive issue is that standard QALY theory treats the outcome of a life-saving operation for a severely disabled patient as worth next to nothing. Intuitively, it is worth a lot more than that. The inequality we object to is the disproportionately low value placed on the outcome for disabled patients. Correcting this unjust valuation does not require treating all outcomes as equally valuable. It just requires us to recognise the true value of the outcome, and give equal weight to equal value. If we succeed in this endeavour, we would expect to see a significant reduction in the number of cases of discrimination. However, discrimination need not be eliminated entirely.
To understand the nature of our pro-equality intuitions, notice the parallel in the situation in which an employer assumes that a job applicant from an ethnic minority lacks a desirable trait, for example, intelligence, without testing the candidate for that trait. We do not object to an employer discriminating according to intelligence, if intelligence is needed for the job. However, we do object to an employer jumping to conclusions about an individual's intelligence based on their race.
In the case of healthcare resource allocation I suggest there is an analogous problem with assuming that the outcome of a life-saving operation for a disabled patient is of lower value than the same outcome for a non-disabled patient, just because that is how the general public values it. It seems to me the general public's valuation of the outcome for disabled patients is ignorant, such that decisions based on that valuation are unfair, in the same way as the employer's attribution of low intelligence to an ethnic minority candidate would be ignorant. The general public's valuation is based on assumptions rather than a more reliable method such as direct experience.
By excluding valuations based on ignorance and by giving a ‘voice’ to disabled patients in these decisions, I contend we respond adequately to the pro-equality intuitions that drove us towards patient preference theory. Yes, there is a residual inequality, but some inequality is not in itself an indictment of a theory. It is unavoidable that some patients will be funded and some will not. However, we must eliminate unfair inequality, and I think patient preference theory does that.
The author thanks Esther Sinclair, Katherine Payne, Riccardo Pingiori, Michael Donnan, Jean Hewitson and members of the audience for their helpful comments at the University of the West of England Concepts of Health Conference, September 2010; also Alasdair Richmond for early encouragement. Finally, thanks to an anonymous reviewer for very helpful suggestions.
Competing interests SCS conducts market research and opinion research for pharmaceutical companies and healthcare providers.
Provenance and peer review Not commissioned; externally peer reviewed.
↵i These public and patient preference figures are assumed for the sake of argument and I do not claim that they represent what we would get from properly run preference surveys, but the public preference figure was guided by a figure quoted in Dolan.18 An EQ-5D public preference survey produced a quality of life figure of 0.152 for health state ‘32211’. In this state, the patient has the worst possible mobility, that is, they are confined to a bed. The patient also has ‘some problems’ with washing and dressing themselves and with performing their usual activities, although they do not have pain or discomfort and they are not anxious or depressed. (Scores for worse conditions are negative, ie, the public thinks they are worse than death, and I have dodged the issue of translating negative figures into quality of life percentages, although lower figures would help my case.)
↵ii The patient preference figure is also very speculative, but inspired mainly by three studies. First, Brickman et al19 report that 29 paraplegic/tetraplegic respondents gave a ‘general happiness’ score of 2.96 versus healthy controls 3.82, on a six-point scale from 0 to 5. Second, Schulz and Decker8 report that 100 paraplegic/tetraplegic respondents gave a mean score for life satisfaction of 10.76 on a scale of 0 to 18, which compared with means for other samples ranging from 12.2 to 13.35. Finally, Bach and Tilton20 report that 42 ventilator-assisted individuals with tetraplegia gave an average score of 4.39 out of 7 for satisfaction with ‘life in general’ compared with 5.36 for healthy controls.
These patients have serious disabilities, yet their average scores for life satisfaction are positive—not as positive as non-disabled comparator groups, yet not as different as you might expect. To be precise, the scores are between 22% and 12% lower than the comparators. Having said that, my line of argument needs a preference-based value, and these studies were not preference surveys. However, with the well-tested survey instruments used in these studies, we would expect positive scores to correlate with positive scores in standard preference-based surveys—otherwise there would be reason to question the validity of these instruments. So in the absence of better data, I hope I do not seem to be wildly overstating my case by supposing that, if we applied a preference-based scale that reflected the value placed by the patients themselves on possible health states, we might find spinal cord-injured patients to have an average 70% health-related quality of life. Another reason for taking this supposition to be modest is that the comparison is with the ‘32211’ health state used for the public preference figure. That state is probably preferable to that of the patients in these studies.
My last line of defence: my argument does not hang on this precise figure. I just need patients' valuations of their own states to be a lot better than public valuations of the same states.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.