Norwegian Institute of Public Health,
December 21 2010
Priority to the worse off: Severity of current illness versus shortfall in life time health.
In evaluating treatment programs for different diagnostic groups, decision makers and the population at large in a number of countries wish to give priority to those with more severe current suffering and/or greater expected future losses of life years and quality of life. Tools have been developed to incorporate such concerns in formal economic evaluation of health programs. On the other hand, concerns for equity in life time health are expressed in comparisons of health in different socioeconomic or geographical groups. Data on the strength of societal preferences for group equity in life time health are at present almost non-existent. To collect such data is an interesting challenge for future research, but it is not necessarily helpful to measure such preferences at a cardinal level in terms of numerical trade-offs.
A common feature of ethical theories of fairness in resource allocation and of the thinking of ordinary people about fairness in health care is that benefits are considered to have greater value the worse off those who receive them are (Rawls, 1971; Daniels, 1993; Nord, 1999). Concerns for the worse off run counter to the utilitarian idea that resources should be allocated simply with a view to maximising overall health benefits. Parfit (1991) refers to concerns for the worse off as ’The Priority View’. Brock (2001) offers a number of possible justifications for the view, including the argument that the worse off suffer undeserved relative deprivation, and/or that the worse off have more urgent needs.
If priority in health care is to be given to the worse off, there is first a question of whether one should be concerned about those worse off in health or those worse off overall, i.e. in their global life situation. While some would argue in favour of the latter, there is probably less agreement about this than about giving priority to those who are worse off in terms of health. In this paper I focus on worse-offness in terms of health. Given this focus, I address a question that has been raised regarding the time perspective of ‘worse-offness’: Is being worse off a matter of being in a bad state at a given point in time, or of having a bad prognosis, or of having had a poor history of health, or perhaps all of these, i.e. of having large health losses as judged over a whole lifetime?
The late British economist
Amartya Sen (2001) regards Williams’ approach as an interesting and potentially powerful one, particularly since it seems to deal with social class inequality in a fulsome way. But he also stresses its limitations for policy making. For example, Williams claims that men are not getting their fair innings, insofar as their health adjusted life expectancy is significantly lower than that of women. While acknowledging the latter fact, Sen suggests that giving preference to male patients ’cannot but lack some quality that we would tend to associate with the process (my italicising) of health equity’ (p.21). Sen thus warns against approaches that insist on taking a single-dimensional view of health equity, stating that ’it is possible to accept the significance of a perspective, without taking that perspective to be ground enough for rejecting other ways of lookiing at health equity, which too can be important’.
My own position is similar to that of Sen. I hold that health can be defined and measured in several meaningful and policy relevant ways, including ‘current’, ‘future’, ‘past’ and ‘life time’ health. Individuals’ scores on these different measures are correlated, partly because illness early in life is empirically correlated with illness later in life, partly because the concept of life time health subsumes past, current and future health. But the correlations are not necessarily high. A person may be worse off than another person an all measures, but he can also be worse off on one measure and at the same time better off on another one. For some allocation problems, decision makers may thus find it helpful to have descriptive information on worse-offness on several or all of the possible health measures. At the same time, ethicists and policy makers may have normative views with respect to the relative weights that should be assigned to different aspects of worse-offness in priority setting between patients and patient groups.
In my own previous work on worse-offness I have focussed on so-called ‘severity of illness’ (Nord, 1993; Nord et al, 1999). This includes current impairments and symptoms and expected future loss of quality of life and/or length of life due to the illness. The main reason for this focus is that official Norwegian guidelines for priority setting in health care, proposed by the Ministry of Health and endorsed by Parliament, since 1987 place great emphasis on severity of illness thus understood (Norwegian Priorities Commission, 1987). By contrast, health losses in the past or aggregate health losses over the whole life time have never since been mentioned as relevant - let alone salient - factors for priority setting in Norwegian policy documents. The same focus on the severity of current and future illness, rather than on past health losses or life time health, is to be found in official guidelines for priority setting in the early 1990s in the Netherlands, New Zealand and Sweden (Dutch Committee on Choices in Health Care, 1992; Campbell and Gillett, 1993; Swedish Priorities Commission, 1993), and in the medical ethics literature (Daniels, 1993). Lately, the Dutch position has been reinforced by a Government guideline saying that willingness to spend public money in order to gain a QALY will range from 10.000 euros for conditions of little severity to 80.000 euros for conditions of great severity (College voor zorgverzekeringen, 2009)
So what do I then think of proposals – put forward for instance by Williams (2001) - to focus more on shortfalls in health over the whole life time and less on the severity of a person’s current and future health problem?
My first answer is that the younger the person or group in question, the less difference is there between focussing on current and future severity and focussing on expected life time health. At birth there is no difference.
On the other hand, in older intervention groups focussing on life time health rather than on severity can make a big difference for the valuation of a given health benefit. For instance, pain acquired at the age of 70 may be more unpleasant than the symptoms of a pollen allergy acquired at the age of 10. But the aggregate QALY loss over the whole life time may be greater in the latter case than in the former. So whom is it more important to treat with a new medicine – the 70-year old with the new pain or the 70 year old with the old allergy?
I return to this question below, But first I want to make a general point. I believe that the relevance of different measures of worse-offness for societal priority setting and willingness to pay for health benefits probably depends on the nature of the decision problem. To see this, consider the two following examples. In problem A, a national medicine administration is to determine its willingness to spend public money on reimbursement of a new and better drug for people with a given chronic illness X, compared to its willingness to pay for reimbursement of a new and better drug for people with a given chronic illness Y. Assume that those with conditions X and Y are of all ages and that the two drugs yield the same health benefit (measured for instance in QALYs). It then seems ethically plausible that the willingness to pay be strongly influenced by the degree of need in the two patient groups - defined as the current and expected future health losses associated with the two illnesses. In problem B, on the other hand, a regional health authority is to determine its willingness to pay for a health education program targeting adults living in an area where life expectancy at birth is 70 years, compared to its willingness to pay for such a program in an area where life expectancy at birth is 80 years. In this case, it seems ethically plausible that the regional health authority’s willingness to pay be higher in the former area on account of that area’s poorer performance on life time health.
Many more examples could be given. But I think these two suffice to demonstrate that more than one measure of health and worse-offness in health deserve a place in the economic evaluation and priority setting tool box.
This said, there are some challenges facing the use of life time health as a statistic for expressing worse-offness.
First, as already touched upon, focussing
on life time health losses may disadvantage elderly people who have lived most
of their life in good health but in old age come to need relief of – perhaps
severe – discomfort. Particularly if the fair innings argument is allowed to
prevail, relief of discomfort in the elderly will – all else equal - be given
less priority than similar relief in younger people (Nord, 2005). I find this
difficult to justify ethically. In publications following his initial 1997
Second, there is in priority setting in national health services a de facto focus on current suffering and expected future health losses. There is thus in a sense an ethical and political burden of proof resting on those who argue in favour of a larger role for the life time health approach in priority setting. I believe they need not to argue on a general basis, but to consider carefully in which specific types of comparisons and decisions it might be useful to supplement current evaluation practice with life time health considerations. As indicated above, the usefulness is probably greater in comparison of preventive programs for different social groups than in comparison of treatment programs for different diagnostic groups.
Third, in specific decision problems where
life time health considerations are deemed salient, there are measurement
issues. One concerns level of measurement. In decisions about priority setting
between technologies and interventions for different diagnostic groups,
representation of population values in cardinal (numerical) terms has proved interesting
and useful for decision makers, cfr NICE’s heavy reliance on utility
measurements and QALYs in the UK and the recent Dutch introduction of varying
limits to willingness to pay for a QALY depending on the severity of the
targeted condition. The Norwegian Medicine Agency, in its continuous dealing
with applications from the pharmaceutical industry for having new drugs listed
for reimbursement, has expressed interest in numerical guidance similar to that
recently introduced in the
If one nonetheless should wish to incorporate concerns for equity in life time health in numerical evaluations of health programs, how should one go about establishing the strength of social preferences for giving priority to groups who are disadvantaged? On this account, research in the past on the strength of concerns for severity may have something to offer.
Table 1. Using the time trade-off (TTO) and the person trade-off (PTO)
to determine weights for life years gained in different social groups.
Social group Life expectancy TTO PTO Weights
1 80 5 30.000 1.0
2 76 2 15.000 1.5
3 72 1 10.000 2.5
Consider table 1. Assume three different social groups (socioeconomic, geographical, ethnic or other) with life expectancies at birth of 80, 76 and 72 years respectively. Appropriate groups of reasonable people could be asked to deliberate about a time trade-off (TTO) question: How many years of increased life expectancy in group 1 and 2 would they deem as equally valuable as an increase by one year of life expectancy in group 3? Median (or mean) answers, or consensus answers, might for example be 5 and 2 years respectively. Similarly, they could be asked to deliberate about a person trade-off (PTO) question: How many persons in group 1 and 2 would have to gain a year of life expectancy for that to be deemed as valuable as 10.000 people in group 3 obtaining such a one year gain? Median answers (or the consensus answer) might for example be 30.000 and 15.000 persons respectively. The ratios expressed in the two sets of responses (TTO and PTO) would most likely not be mathematically consistent with each other, since different framings of problems are known to lead to different answers (Kahneman and Tversky, 1979). But policy makers could take both sets of responses into account and make a rough overall judgement. They might thus decide that in formal economic evaluation of programs that reduce mortality risks in different social groups, a set of weights for life years as those in the right hand column of tablee 1 might be used in order to roughly capture societal concerns for social group equity in life expectancy. In principle, concerns for social group inequalities in quality adjusted life expectancy, as measured in terms of QALYs or DALYs, can be accommodated in the same way.
Weighting life years or QALYs or DALY gained in different social groups is a rather simple way to incorporate concerns for equity in life time health in formal program evaluation. A more sophisticated approach is to use a societal valuation function of the form SV = Qa x Eb, where Q is the total number of QALYs gained by a program and E is the resulting degree of equality in life time health across individuals (Wagstaff, 1991; Norheim, 2010). The values of the parameters ‘a’ and ‘b’ are set such as to reflect the importance that societal decision makers place on total QALY gains relative to equality in life time health. The parameters thus express a trade-off between these two concerns.
One challenge in this approach is to decide who should be included in the measurement of equality. For instance, if a program targets a population subgroup A, what is the total population in which the resulting degree of equality in life time QALYs should be measured? It can hardly be the whole population, since few programs are big enough to have a noticeable effect on equality in the whole population. Should it be therefore be subgroup A plus another subgroup B which is targeted by a competing program? But then, why particularly program B rather than C or D or E? Presumably one would want to look at a set of programs and value each of them in turn in the light of resulting QALY gains and effects on equality measured across all the target groups in question. The measurement of such equality effects is no small task.
Another challenge is to estimate the
parameters ‘a’ and ‘b’ through preference elicitations. This can be
complicated. For instance, in a study for NICE in the
Figure 1. Example of question about the trade-off between aggregate health and its distribution.
Scenario X Group 1: 60 years in full health, 8 years in poor health.
Group 2: 56 years in full health, 8 years in poor health.
Scenario Y Group 1: 72 years in full health, 16 years in poor health.
Group 2: 48 years in full health, 16 years in poor health.
I conclude that in evaluating treatment programs for different diagnostic groups, decision makers and the population at large in a number of countries wish to give priority to those with more severe current suffering and/or greater expected future health losses. Representation of population values in cardinal (numerical) terms seems to be of interest to government bodies working on a regular basis with technology assessment and priority setting across diagnostic groups. By contrast, concerns for equity in life time health are mostly expressed in comparisons of health in different socioeconomic or geographical groups. Data on the strength of societal preferences for group equity in life time health are at present almost non-existent. To collect such data is an interesting challenge for future research. However, it is not clear that decision makers will find it helpful to have such preference data at a cardinal level of measurement.
Acknowledgement: I am grateful to Lars Granum and Morten Aaserud at the Norwegian Medicine Agency for earlier exchanges on which I have drawn in this paper and to Einar Anders Torkilseng at the Norwegian Directorate for Health for having provided useful comments to a draft version of the paper.
Brock, Dan. ‘Priority to the worst off in
health care resource prioritisation.’ In Battin M, Rhodes R, Silvers A (eds).
Health Care and Social Justice.
College voor zorgverzekeringen. ‘Pakketbeheer in de Praktijk
Dutch Committee on Choices in Health
Care. ‘Choices in Health Care.’
societal concerns for fairness in numerical valuations of health programs.’ Health Economics, 8 (1999):25-39.
distributive justice.’ International Journal for Equity in Health, 8 (2009):40 doi:10.1186/1475-9276-8-40.
Norwegian Commission for Prioritising in
Health Care. ‘Retningslinjer for prioritering innen helsevesenet. (Guidelines
for prioritising in health care.)’ NOU
Parfit, Derek. ‘Equality or priority.’ The Lindley Lecture. 1991. Copyright:
Department of Philosophy,
Rawls, John. ‘A theory of justice.’
Sen, Amartya. ‘Why health equity.’ Keynote address to IHEA conference,
Slovic, Paul. ‘The construction of preference.’ American Psychologist, 50 (1995):364-371.
Swedish Health Care
and Medical Priorities Commission. ‘No easy choices - the difficulties of
health care.’ SOU 1993:93.
Williams, Alan. ‘Intergenerational equity: an exploration of the "fair innings" argument.’ Health Economics, 6 (1997):117-132.