Introduction
The UK is an unequal society. If you are poor, you can expect to live a shorter life than if you were rich [
1,
2], you can expect to live with lower average health-related quality of life [
3], and you can expect to experience disability at a younger age [
4]. This “health gap” is substantial [
5]. In quality-adjusted life-year (QALY) terms, a person living in the most deprived quintile areas of English society can expect to experience 11.87 QALYs less in their lifetime than a person living in the least deprived areas [
3].
Recent evidence suggests the UK public are averse to this inequality, and would be willing to sacrifice a significant amount of average population lifetime health to achieve a more even distribution of it between socioeconomic groups
1 [
6‐
9]—they appear to be “distributionally sensitive”. In contrast, economic evaluation in health is typically “distributionally naïve” [
10], and operates under the assumption that “a QALY is a QALY is a QALY” [
11], irrespective of who receives it. This apparent discordance has led some to question the democratic legitimacy of distributionally naïve approaches, and to call for distributionally sensitive forms of economic evaluation, such as “distributional cost-effectiveness analysis” [
12‐
14].
If the UK public’s preferences towards inequalities in health are to be captured in distributionally sensitive economic evaluation, it would be valuable to understand the answers to three questions: (1) How averse are the UK public towards inequalities in lifetime health between socioeconomic groups? (2) Does the extent of that aversion differ depending upon the type of health (e.g. life extension, pain relief or mobility improvement) under consideration? (3) Are the UK public as averse to inequalities in health between socioeconomic groups as they are to inequalities in health between neutrally framed groups? This third question matters, as it is not immediately obvious whether or not it is normatively desirable for social health-related resource allocation decisions to be made based on socioeconomic status, or whether they should be based on health alone [
15]. This systematic review focuses on these three questions.
Previous systematic reviews have focused on general public preferences regarding different broad criteria for prioritisation [
16,
17], or preferences regarding differences in the future health of individuals [
18‐
20]. This is the first systematic review to focus explicitly on the UK public’s aversion to inequalities in lifetime health between socioeconomic groups, although an unsystematic review has recently been published [
21]. The scope of this review is restricted to the preferences of the public in the UK, as the primary objective of the study is to inform distributionally sensitive economic evaluation in the UK.
Methods
Search strategy
Four databases were searched: Ovid MEDLINE (1946—27/10/2017), Ovid EMBASE (1974—26/10/2017), Ovid EconLit (1886—30/09/2017), and Web of Science’s Social Sciences Citation Index (SSCI) (1956—27/10/2017). All searches were undertaken on 27/10/2017.
The search strategy was developed in an iterative fashion. First, six “pearls” [
22] were identified as starting points, to provide the initial list of key words [
6,
8,
23‐
26]. Second, the MeSH headings associated with these papers were recorded, and a word frequency analysis of the paper titles/abstracts was undertaken [
27]. These were supplemented with additional terms based upon the search questions to generate an initial search strategy.
2 Following this, the reference lists of the pearls were reviewed, to identify additional papers. The sensitivity of the draft search strategy was then tested in MEDLINE, by assessing whether or not it could return the papers identified from those reference lists. If a paper was not identified, the search strategy was then updated with key terms from the unidentified paper. Further scoping searches were then conducted based upon this revised strategy, and the reference lists of potentially relevant papers scanned for other potentially relevant papers.
The search strategy was then tested again to assess whether it identified all papers identified in scoping searches, refined as needed, and the same process repeated until the reference list of all papers identified in scoping searches were picked up by the search strategy. The final MEDLINE search strategy is detailed in Online Appendix 1. Following the screening of the database search results, the selected papers were reviewed in detail, to identify potentially relevant journal publications, or grey literature, not captured within this search. These papers were then treated as new records, and screened accordingly.
Eligibility criteria
Papers were assessed for eligibility using six hierarchical inclusion criteria. First, papers published in English were included, and all others were excluded. Second, publications in peer-reviewed journals, reports published by NGOs/HTA bodies, and studies published in discussion papers by academic institutions, were included. All other publication types, including conference abstracts, were excluded. Third, experimental studies in which the stated preferences of participants were quantitatively elicited were included. Non-experimental revealed preference studies, non-quantitative studies, and reviews of prior studies, were excluded. Fourth, those studies featuring broadly representative samples of the UK adult general public were included.
3 Studies centred on selective samples of the UK population, such as students, policy makers and health care professionals were excluded. Studies featuring exclusively non-UK respondents, or for which it was not possible to isolate the preferences of UK respondents, were excluded. Fifth, studies were assessed for their ability to provide information on the extent of the public’s aversion to inequalities in lifetime health between socioeconomic groups. Studies that explicitly asked, or could be implied as asking, respondents to make efficiency/equality trade-offs between individuals, or groups, with differing lifetime health in a range relevant to socioeconomic inequalities in health (life expectancy
4 or quality-adjusted life expectancy: > 50 and < 90 [
1]) were included—irrespective of whether participants were told they were choosing between socioeconomic groups, or between neutrally framed groups in a comparable range of lifetime health. Two distinct strands of empirical literature were considered to be capable of providing this information—(1) stated preference studies focused on health inequality aversion
5 [
28‐
30], and (2) stated preference studies focused on eliciting preferences regarding prioritising those individuals with a higher Burden of Illness, as defined by their absolute QALY shortfall in prospective health attributable to some illness
6 [
31,
32].
Studies that did not apply a lifetime time-horizon, or that could not be utilised to infer aversion to lifetime health, were excluded. Stated preference studies that focused on severity, as defined by relatively poor quality of life [
19], and preferences regarding treatment at the end of life [
18] were excluded for this reason. Studies focused explicitly on inequality aversion in the context of gender, or differences in lifestyle, were also excluded. Finally, the choice perspective employed in each study was evaluated. Those studies that asked respondents to make choices in the context of public resource allocation decisions that did not affect them personally, for example how to allocate finite NHS resources between two groups they were not part of, were included. Those studies that asked respondents to make choices that would impact them, for example their willingness to trade-away their own wealth, were excluded.
Study selection
Study selection was conducted using a two-step process, with titles and abstracts screened first followed by screening of full papers. Eligibility criteria were applied sequentially in the order detailed above, with the first arising reason for exclusion recorded. The first two waves of screening were conducted by Simon McNamara. Abigail Steveley then independently reviewed a random sample of 20 full papers against the eligibility criteria. This independent review identified one discrepancy: the decision of whether or not to include a study by Petrou et al. [
33]. The lead author of the study was contacted to clarify whether it used a general population sample, which resolved the discrepancy and the study was included. The audit identified no significant concerns regarding the screening undertaken.
Discussion
This review set out to do three things. First, to identify estimates of the strength of the UK public’s aversion to inequalities in lifetime health between socioeconomic groups. Second, to explore whether the strength of this aversion differs depending upon the type of health under consideration. Third, to explore whether or not aversion differs depending upon whether participants were told that the inequality existed between socioeconomic groups, or neutrally framed groups. We identified 15 studies relevant to these aims.
The identified studies provide general, although not universal, support for the idea that the UK public are averse to inequalities in life expectancy (at birth) between socioeconomic groups. Similarly, the studies identified provide evidence that the public are averse to inequalities in life-expectancy (at birth) between neutrally framed groups in a comparable range of lifetime health. Eleven of the 15 studies identified provide evidence in support of aversion to inequalities in total life expectancy [
6‐
9,
23,
24,
31,
33,
35,
38,
41], two provide evidence in opposition [
36,
39], and two are inconclusive [
37,
40]. However, the strength of aversion differed substantially between studies, with higher levels of aversion elicited for inequalities presented as being between socioeconomic groups than between neutrally framed groups. For example, Petrou et al. [
33] and Baker et al. [
35] estimate relative weights of only 1.37 and 1.55, respectively, on an incremental health gain provided to someone who will die at 60, compared to someone who will die at 80. In contrast, Dolan and Tsuchiya [
8] estimate weights of 6.8–9.95 for a marginal health gain provided to an individual of lower socioeconomic status with a life-expectancy of 73 compared to an individual of higher socioeconomic status with a life-expectancy of 78. Similarly, it is notable that the Atkinson inequality aversion parameters estimated by Edlin et al. [
24] in a neutral context are substantially lower than those estimated by Robson et al. [
6], Ali et al. [
7], and Dolan and Tsuchiya [
8] in a socioeconomic context; see Table
3.
A small number of the identified studies suggest that the public may be more averse to an inequality of a given QALY magnitude if that inequality is due to differences in life-expectancy, rather than quality of life. Both Rowen et al. [
31] and Dolan and Tsuchiya [
23] find that, whilst the public are averse to inequalities in quality-adjusted life expectancy (QALE) attributable to differences in life-expectancy, they are not averse [
31], or as averse [
23], to inequalities in QALE attributable to differences in quality of life. Similarly, in their profile tests, Baker et al. [
35] find that the public prefer to prioritise those with better, rather than worse, past quality of life. This evidence suggests that public preferences regarding inequalities in health may be consistent with the “fair innings” argument based on duration of life, but may not be consistent with the “extended fair innings” argument that adjusts for quality of life [
30]. None of the studies identified explored the possibility that health inequality aversion might depend upon the specific type of health gain under consideration (e.g., comparing aversion in the context of pain relief and life extension).
This review has three primary limitations. First, our inclusion of studies focused on BOI under the assumption that the impact of BOI on preferences is linear, which is an assumption that may or may not hold [
31]. Sensitivity analysis indicates that the exclusion of the two BOI studies identified would not have an impact upon our conclusions regarding aversion to life-expectancy at birth. However, one of the three studies that suggests aversion to inequalities in lifetime health attributable to differences in quality of life may be lower than to those attributable to differences in length of life was a BOI-based study [
31]. As a result, the strength of this conclusion would be weakened by excluding these studies. Second, our search was designed to inform distributionally sensitive economic evaluations conducted in the UK, and so was restricted to evidence on the views of people in the UK. As a consequence, the results themselves may be of limited generalisability to other countries. Third, the studies identified are methodologically heterogeneous, and report estimates of aversion in different ways. This makes it challenging to compare across studies and, with the exception of the four studies for which we calculated Atkinson inequality aversion parameters, it prevents any attempt at formal synthesis. The primary strength of this paper is the fact that it is the first systematic review of this kind; notably, we identified more studies than found in a recent unsystematic review of health inequality aversion [
21].
Four key issues
Our findings raise four issues. First, if the public are averse to inequalities in health, does it make sense to continue to conduct, and use, distributionally naïve economic evaluations? [
43]. Whilst this review demonstrates that it is challenging to quantify precisely how averse the public are to inequalities in health, the evidence available does suggest they are averse. The distribution of health gains appears to matter to the UK public, and ignoring this preference by continuing to conduct distributionally naïve economic evaluation is a choice that runs counter to this preference. Second, if we want to introduce consideration of inequalities into economic evaluation, what level, or levels, of aversion should be implemented in practice?
19 This is a critical question, because the prioritisation of equality has a human cost [
30,
44]. If we choose to prioritise equality, we accept there will be more suffering, and loss of life, than might otherwise be present in our society. Conversely, if we choose not to prioritize equality, we choose to accept that the social burden of ill health will be disproportionately placed on the poor. The level of inequality aversion incorporated in an economic evaluation would quantify the acceptable human cost of a given improvement in equality, and so it is critical to define it in a considered way. This review found wide variation in estimates of public preferences regarding inequalities across studies, which highlights the challenge of selecting a single estimate of aversion to implement. Given this variation, those conducting economic evaluations would be wise to undertake sensitivity analyses surrounding the relative weight they give to the distribution of health gains and average population health gains. If distributionally sensitive economic evaluation is to become more widespread in the UK, it would be valuable for a body like NICE or Public Health England to define a reference level of health inequality aversion (perhaps using a Citizen’s Council comparable to [
40]), so that those conducting these analyses can present their work in a comparable and consistent manner. Again, note that if these bodies do not comment on this issue, this equates to an endorsement of a status quo in which the reduction of inequalities in health carries no weight in economic evaluation. Third, if aversion to socioeconomic inequalities in health is higher than aversion to neutrally framed inequalities of equivalent magnitude, which (if either) strength of aversion is the appropriate one to reflect in distributionally sensitive economic evaluation? Should estimates of aversion from neutrally framed studies be used because this removes the influence of non-health factors upon respondents’ preferences? Or should estimates of aversion from socioeconomically framed studies be used because this reflects the fact that inequalities in health between socioeconomic groups are systematic, as opposed to being random variation within the population, and so may be considered inequitable? Fourth, is health inequality aversion consistent with the QALY model, or does the type of health matter to the public? If aversion does differ depending upon whether the public are asked about life-expectancy, pain relief, or any other form of health gain: how should this be accounted for in distributionally sensitive economic evaluation? Can QALY-based distributional cost-effectiveness analysis represent the views of the public?
In conclusion, this review suggests that the UK public are averse to inequalities in life expectancy between socioeconomic groups, albeit with wide variation in the strength of this preference between studies. We find evidence of aversion between neutrally framed groups; however, the UK public appears to be more averse to inequalities in health between socioeconomic groups. We find limited evidence that the composition of an inequality may impact the strength of aversion, and in particular, that the public may be less averse to an inequality of a given QALY magnitude if that inequality is due to differences in quality of life, rather than life-expectancy.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.