Investigating public preferences on ‘severity of health’ as a relevant condition for setting healthcare priorities

Social Science & Medicine 68 (2009) 2247–2255

Contents lists available at ScienceDirect

Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed

Investigating public preferences on ‘severity of health’ as a relevant condition for setting healthcare priorities Colin Green* Institute of Health Service Research, Peninsula Medical School, University of Exeter, Barrack Road, Exeter EX1 5DW, UK

a r t i c l e i n f o

a b s t r a c t

Article history: Available online 4 May 2009

This study examines the preferences of a sample of the UK general public over the allocation of healthcare resources. Preferences were elicited against scenarios where alternative patient groups are competing for limited resources. Respondents were asked to make a choice between either (i) groups described according to alternative descriptions for severity of health condition, or (ii) groups described according to a broader level of disadvantage (e.g. family income). The survey used a random-location quota sampling approach, and face-to-face interview techniques. Interviews were completed with 261 people in the Southampton area of England. Results showed that the majority or respondents wanted to divide resources equally between competing groups, giving at least equal preference to the more severely affected group, and the more disadvantaged group, regardless of a stated lower potential health gain in these groups compared to alternatives. In the severity of health question 60% indicated that a unit of health gain in a severely affected patient group was of greater social value to that same unit of health gain in a moderately affected patient group, all else equal. When described by level of disadvantage, 80% of respondents stated such a preference, which indicates that they attach a greater social value to a unit of health gain in a disadvantaged patient group, compared to a more advantaged group, all else equal. When given an option to ‘opt out’ of a difficult decision, and to ‘let others choose’, very few respondents (5%) took that option, suggesting that the most common stated preference of dividing resource equally between groups may be a true preference, rather than respondents avoiding difficult decisions. When interpreting the findings from the survey, results suggest that preferences reported to give priority to those more severely affected by their health, may also be a reflection of a broader preference to treat those groups classed as worse-off, in empirical studies. Results are discussed against the growing importance of the empirical ethics literature, and the growing needs of health policy makers to seek out an empirical basis upon which to consider the challenges of setting priorities in healthcare. Ó 2009 Elsevier Ltd. All rights reserved.

Keywords: UK Priority setting Distributive preferences Public preference Severity of health Methodology

Introduction When setting health policy or planning for the delivery of health care, at a national or devolved decision-making level, there is a need to set priorities when allocating resources. Those involved in making such difficult priority-setting decisions have increasingly explored the notions of fairness and equity to inform the decision making process (e.g. PBAC, 2004; Rawlins & Culyer, 2004; NICE, 2005). Yet there remains considerable ambiguity, and a lack of operational guidance, on what might be meant by fairness and equity (Newdick, 2005). Definitions against these distributive ideals are generally given at an abstract theoretical level (e.g. vertical and horizontal equity), or they are commonly confused

* Tel.: þ44 (0)1392 406915. E-mail address: [email protected] 0277-9536/$ – see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2009.03.020

with aspirational notions of equality (e.g. equality of health, equality of access according to need) (Sassi, Le Grande, & Archard, 2001; Sen, 2002; Hauck, Smith & Goddard, 2003). One source of guidance, when setting priorities, is the growing literature on the incorporation of distributive preferences when making resource allocation decisions (e.g. whether to fund technology X for patient group Y), especially the empirical literature that examines a range of potential social considerations. This literature includes empirical research to assess, and often quantify, the importance of factors such as patient characteristics, characteristics of health care interventions, and broader societal considerations (e.g. reducing inequalities), when making resource allocation decisions. This type of research has been referred to as ‘empirical ethics’ (Culyer, 2001; Richardson, 2002; Richardson & McKie, 2005), and there have been a number of informative reviews (Sassi, Archard, & Le Grande, 2001; Schwappach, 2002; Dolan, Shaw, Tsuchiya, & Williams, 2005; Green, 2007).

2248

C. Green / Social Science & Medicine 68 (2009) 2247–2255

In the empirical ethics literature, studies are mostly considering one particular concern for fairness at a time, and contrasting it with prominent efficiency objectives related to cost-effectiveness, and the maximisation of quality-adjusted life-years (QALYs). One such concern for fairness is the consideration of the severity of a health condition (pre-treatment health state). The empirical studies informing the severity of health approach (e.g. Nord, 1993a,b; Nord, Richardson, Street, Kuhse, & Singer, 1995; Dolan, 1998; Ubel, 1999; Cookson & Dolan, 1999; Oddsson, 2003; Gyrd-Hansen, 2004), have been widely cited in the health care literature to support a hypothesis that health state values (e.g. QALYs) often fail to capture and represent the nature of public preferences. But, more specifically, they have been used to support a ‘severity hypothesis’ where the basic premise is that the social value of a health improvement (of a given size) is greater the greater the severity of the initial health condition (Nord, 2005). The severity hypothesis may be a fair reflection of public preferences, and a reasonable approach for setting priorities. But current empirical research in this area, as with almost all notions of distributive preference, comprises studies that are mostly simple experimental studies, often without clear policy implications, and further research is required against a severity of health criterion in the context of a broader range of distributive preferences. In this paper, as well as considering the empirical support for the use of severity of health as a priority-setting criterion, a further aim is to explore the broader interpretation of the empirical evidence reporting a stated preference to treat those groups regarded as more severely affected by their health status. This latter aim is informed by a recent systematic review of the literature around fairness and equity considerations, in health care prioritysetting dilemmas (Green, 2007), covering over 100 empirical studies. This review, in a very diverse and difficult to interpret literature, noted that in many studies respondents favoured the worst-off groups (e.g. those with worse health condition, poorer prognosis, those subject to greater inequality), from a set of groups competing for the same scarce resources. This observation was visible across a range of different study settings (contexts), and in studies employing a range of methodological approaches, with those groups that would be regarded as ‘worst-off’ being given equal or priority-preferences, regardless of their more limited capacity to benefit (e.g. lower health gains). This notion of ‘worseoffness’ has been referred to in the literature (Nord, 2005), but has not been considered directly, in the context of preferences around treatment of those more severely affected by their health. Therefore, one hypothesis explored here is that a preference registered for the more severe of two competing groups in an allocation problem, reflects a preference to treat ‘the worst-off group’. With such a preference reflecting not only a social value to treat severely affected groups (in specific studies), but a broader and general preference for ‘fairness’ in the treatment of worse-off groups. The motivation here is not to disprove the specific hypothesis of a preference for severity in such instances, but to further investigate the meaning of the preferences reported to support a severity of health argument. Severity of health as a basis for setting healthcare priorities The empirical ethics literature does provide support for the potential use of severity of health (severity of illness) as a basis for informing healthcare priorities (see above). For example, studies by Nord (1993a) and Ubel (1999) have suggested respondents have a preference to prioritise either equally or in favour of the most severely ill patient groups, regardless of the potential efficiency gains (health output) forgone. These two commonly cited studies are used in the current paper as a point of departure

(methodological framework) for further research on severity of health as a relevant condition for setting priorities. Nord (1993a) reports a study (published in Norwegian, but reported in Nord, 1999), to elicit the preferences of Norwegian health policy planners. He presented subjects with the following scenario: Imagine an illness A that gives severe health problems and an illness B that gives moderate problems. Treatment will help patients with illness A a little, while it will help patients with illness B considerably. The cost of treatment is the same in both cases. There is insufficient treatment capacity for both illnesses and an increase in funding is suggested. Three different views are then conceivable. (1) Most of the increase should be allocated to treatment for illness B, since the effects of these are greater, (2) Most of the increase should be allocated to treatments for illness A, since these patients are more severely ill, (3) The increase should be divided evenly between the two groups. Which of these views comes closest to your own? The results reported show that very few respondents favoured moderately ill patients, most preferred to divide resources equally, the remainder favoured treating the severely ill. The format used by Nord was a simple one, and Ubel (1999) sought to examine how stable the preference for treating the severely ill would be when the wording used by Nord to elicit preferences was modified. The Ubel study included the original question framed by Nord, and the same question using modified text (see below), with these questions having three response options (i.e. allocate resources to A, or B, or to divide evenly), as in Nord (1993a). The focus in the current study is on the preferences contrasted by Ubel with those of Nord’s earlier study (Nord, 1993a), and development of this simple study design. In the Ubel study the introductory text was the same as above, with modifications made to the Nord (1993a) text used to describe the response options. These alterations comprised subtle, but important, text differences, with respondents asked which of the following three views (responses) came closest to their own: (1) Most of the increase should be allocated to treatment for illness B, involving moderate health problems which improve considerably with treatment (2) Most of the increase should be allocated to treatment for illness A, involving severe health problems which improve a little with treatment (3) The increase should be divided evenly between the two groups The text in the response options used by Ubel no longer contained suggestions of conflicting rationales, and text is arguably more independently descriptive. Ubel used a USA sample of the general public, compared to Nord’s sample of Norwegian politicians, yet Ubel reports that for the replication of the Nord study (Nord text) results were similar. Very few respondents favoured moderately ill patients (9%), most preferred to divide resources equally (64%), the remainder favoured treating the severely ill (26%). Where text is modified by Ubel (1999) he reports similar results to those of Nord, in so far as the majority of respondents were willing to give at least equal priority to the severely ill patient group, regardless of the forgone health benefits. But only 6% of respondents expressed a preference in favour of the more severely ill patient group (Group A). Ubel reports that 21% of respondents favoured moderately ill patients (Group B), and that most (73%) preferred to divide resources equally. Ubel offers some considered discussion on the methods used, and on the results presented in the context of the general literature. Ubel concludes that the study ‘‘showed that public preferences for

C. Green / Social Science & Medicine 68 (2009) 2247–2255

treating severely ill patients are not as strong as suggested by Nord’s study. However, the data still suggest that many people prefer helping severely ill patients, even when they benefit significantly less than moderately ill patients; almost half of the subjects favoured treatment for the severely ill patients despite this large difference in treatment benefit,’’ (Ubel, 1999, p. 902). Ubel states his study has no immediate policy implications, however findings are commonly cited (e.g. Schwappach, 2002; Dolan et al., 2005) to support ‘severity of health’ as a priority setting (preference) criterion. Whilst these studies offer support for the severity hypothesis, there are some uncertainties in the confidence we can place in the results presented. The methods used are simple self-complete questionnaires, in convenience or purposive samples. The studies are experimental in nature, with methodological motivations rather than being aimed at influencing health policy. The nature of the preferences presented, with a large proportion of respondents choosing to divide resources equally between the two groups, also urges caution in drawing conclusions that the preference is indeed an accurate one, and not just an artefact of the choices available i.e. respondents indicating an equal division of resources as they were unwilling to make a difficult treatment choice. Oddsson (2003) has added to the empirical evidence reported by Nord (1993a) and Ubel (1999), using a similar questionnaire format, but the interpretation of the findings against severity of health are confounded by the presence of other decision-making influences. Methods: Empirical study This study examines the preferences of a sample of the UK general public in the allocation of healthcare resources, where groups competing for limited resources are described using differences in severity of health, and alternatively differences in level of disadvantage. The aims of the study were: (a) to add to the empirical evidence on the role of severity of health in the priority setting debate, using a sample of the general public; (b) to explore

2249

a general hypothesis that a preference to treat those more severely affected by their health also represents a general preference in favour of the ‘worst-off’ groups in society; and (c) to consider the meaning of a choice to give equal preference to competing groups, in empirical studies. Survey instrument The survey used four differing questionnaire formats (Q1–Q4) to address the research questions posed. These formats are summarised below, and presented in more detail in Table 1.  Q1: replicated the question used by Ubel, 1999 (modified version of Nord, 1993)  Q2: was identical to Q1 but with the addition of a fourth response category giving the respondent the opportunity to state that they were not able to make a decision and would prefer that the choice be made by others  Q3: used the same approach as Q1, but replaced the terminology for severely and moderately ill with ‘disadvantaged patient group’ and ‘more advantaged patient group’  Q4: was identical to Q3 but with the addition of a fourth response category giving the respondent the opportunity to state that they were not able to make a decision and would prefer that the choice be made by others Respondents were presented with only one of these prioritysetting questions, containing two competing scenarios, and they were asked to indicate a preference from the response categories available. Interviewers read out the question, and the range of response options, in full, and showcards were used to provide summary information. Appendix A provides an example of the showcards used. In questions three and four, the respondents were informed that both patient groups have similar health problems without treatment. The survey design was driven by the methods used in earlier studies by Nord (1993a) and Ubel (1999), but underwent a series of

Table 1 Format used for the interview questions (Q1–Q4). Q1–Q2: Severe versus Moderate Imagine an illnessdillness Adthat gives severe health problems, and an illnessdillness Bdthat gives moderate health problems. Treatment will help patients with illness A a little, while it will help patients with illness B considerably. The cost of treatment is the same in both cases. An increase in funding is available but we are unable to treat both patient groups. Which of these three (four) views come closest to your own? Q1 (1) Most of the increase should be allocated to treatment for illness B, involving moderate health problems which improve considerably with treatment (2) Most of the increase should be allocated to treatment for illness A, involving severe health problems which improve a little with treatment (3) The increase should be divided evenly between the two groups

Q2 (1) Most of the increase should be allocated to treatment for illness B, involving moderate health problems which improve considerably with treatment (2) Most of the increase should be allocated to treatment for illness A, involving severe health problems which improve a little with treatment (3) The increase should be divided evenly between the two groups (4) I am not able to make a decision and would prefer that the choice be made by others

Q3–Q4: Disadvantaged versus More Advantaged Imagine an illnessdillness Adwhere the patient group is disadvantaged, for example, from a low income family, and an illnessdillness Bdwhere the patient is from a more advantaged group. Treatment will help patients with illness A a little, while it will help patients with illness B considerably. The cost of treatment is the same in both cases. An increase in funding is available but we are unable to treat both patient groups. Which of these three (four) views come closest to your own? Q3 (1) Most of the increase should be allocated to treatment for illness B, in a more advantaged patient group, which improves patients considerably with treatment (2) Most of the increase should be allocated to treatment for illness A, in a disadvantaged patient group, which improves patients a little with treatment (3) The increase should be divided evenly between the two groups

Q4 (1) Most of the increase should be allocated to treatment for illness B, in a more advantaged patient group, which improves patients considerably with treatment (2) Most of the increase should be allocated to treatment for illness A, in a disadvantaged patient group, which improves patients a little with treatment (3) The increase should be divided evenly between the two groups (4) I am not able to make a decision and would prefer that the choice be made by others

2250

C. Green / Social Science & Medicine 68 (2009) 2247–2255

pre-pilot and formal pilot stages, to assess the level of understanding, and acceptability, in a general population sample. The use of the terms ‘disadvantaged’ and ‘more advantaged’ to change the labelling of the worst off group in the scenarios for Q3 and Q4, is based on the common use of such terminology in the social science literature, to reflect a range of possible disadvantages and/or social inequalities (e.g. UK Office for National Statistics, 2002, 2004). In the UK, The Office for National Statistics (ONS) defines disadvantaged households in a number of different ways (e.g. worklessness, lower income, lone parent households, households with state benefits as only source of income, households with no adults born in Britain), (ONS, 2002). In the survey instrument, lower income was used as an example of a disadvantaged group i.e. disadvantaged group (e.g. lower income family). This followed a series of informal pre-pilot interviews, and a formal pilot study of 25 interviewees, where findings supported the acceptability of the terminology. Pre-pilot research investigated other ways to describe two groups where one group was clearly worse off than the other. For example, clinical prognosis (good or poor), length of lifeexpectancy, and different categories of social class, were all considered as potential means for describing the two groups in the survey. The judgement was to use different presentations of disadvantage in the current survey, to represent the broader notion of ‘worse-offness’ (Nord, 2005), which was the primary focus. Sample The main survey instrument was administered on a sample of the adult general population (aged 18 and over) in the Southampton (UK) City Council area. Interviews were carried out in a face-to-face format in-home by the Ipsos MORI Social Research Institute during September and October 2005. The adopted approach to sampling was a random location quota sample. This sets fixed quotas of people to be interviewed in a number of randomly selected sampling points. Sampling points were based on ‘Output Areas’ (OAs) in the Southampton City Council area, the smallest building block of the Census. For the sample used in this survey 32 OAs were randomly selected (by MORI) proportionate to population size, controlling for sociodemographic composition. Quotas were set by MORIdindividually at each sample pointdto reflect the socio-demographic profile of residents, on gender, age and work status, using profile data from the 2001 Census (Southampton City Council, 2004). Assignment of questionnaire Two versions of the main survey instrument were used, each contained two of the four priority setting questions, i.e. the above question format Q1–Q4. Version 1 of the survey instrument contained Q1 and Q3, version 2 contained Q2 and Q4. Respondents were randomly assigned one of the two main survey instruments and interviewers rotated the questions (Q1 or Q3/Q2 or Q4) in subsequent interviewees. Respondents answered only one of the questions Q1 to Q4. This question was presented to respondents, as the second part of a two-part priority-setting survey. The first part of the survey elicited preferences over priority setting scenarios, and methods and results are reported elsewhere (Green, 2009, Health Economics, in press). Sample size The survey aimed to interview 250 people. There are no formal sample size calculation methods for this type of public preference

study. The use of face-to-face interview methods was a limiting factor in the sample size used here, and the target sample size was determined on the basis of available time and resources. Overall 250 respondents was judged to be a reasonable target for a study of this nature, when considering the literature generally. The study by Ubel (1999) reports findings from sub-groups of respondents circa. 70 per question frame, therefore the use of a total sample size of 250, across the four frames Q1–Q4, was thought to be acceptable, given that there was some scope for combining subgroups (Q1–Q4) in the analyses. Results Interviews were completed by 261 respondents. The sociodemographic characteristics of the sample are presented in Table 2. The sample is generally representative of the population in the Southampton City area, being similar in gender, age profile, work status, ethnicity, home ownership, experience of illness/disability, and health status in general (Southampton City Council, 2004). However, the sample includes a higher proportion of retired households, part-time workers, and home workers than in the general population, as might be expected of a face-to-face ‘inhome’ survey of this kind. Table 3 presents sample characteristics by questionnaire version/group (Q1–Q4), and shows that characteristics did not differ by questionnaire version.

Table 2 Socio-demographic characteristics of the sample used.

Total Gender Male Female Age 18–34 35–54 55þ Social grade AB C1 C2 DE Work Status Working full time Not full time Working part-time House person Retired Registered unemployed Unemployed (not registered) Permanently sick/disabled Student Household income Below £17,500 £17,500–29,999 £30,000–49,999 Above £50,000 Ethnicity White British BMEa/other Home ownership Owner occupier Social renter Private renter Illness/disability Yesb No Household composition With children Without children a b

No. of people

%

261

100

116 145

44 56

91 92 80

35 35 30

36 71 65 85

14 27 25 32

94 169 43 23 60 5 4 14 20

36 64 16 9 23 2 2 5 8

92 47 39 16

35 18 15 6

221 40

84 15

150 60 47

57 23 18

103 152

39 58

87 171

33 65

Black and minority ethnic. Respondent and/or someone else in household.

C. Green / Social Science & Medicine 68 (2009) 2247–2255

Table 4 reports response data against choices for Q1–Q4. There was a high rate of engagement in the survey, with 251 of 261 respondents providing a preference when presented with the survey question. Indications from interviewers were that respondents were interested in this type of health related survey question and were enthusiastic about participating. There were only 10 (3.8%) respondents who did not provide a preference across the response categories presented, indicating ‘don’t know’ (or other) when asked to make a choice. Where the study offered respondents an opportunity to avoid making a difficult decision and to let others choose (Q2 and Q4), very few respondents selected the ‘others to choose’ option, only between 3.1% and 7.4% (7/132 people in total, 5.3% when pooled). This indicates that where respondents are stating a preference for equality it may be interpreted as a true preference to divide resource equally between the two groups. Whilst there are only a small number of cases (10 in total) where respondents have not provided a preference (i.e. don’t know or other), the numbers are

2251

smaller in the questions where respondents were able to opt to ‘let others choose’ (Q2 and Q4). The most frequent response was to divide resources equally between groups A and B, as seen with previously reported studies (Nord, 1993a; Ubel, 1999). Where Q1 and Q2 used the format of Ubel (1999), using the modified text around severity of health, the results here do differ from those of Ubel. A larger proportion of patients, around 30%, preferred to direct resources to treat group B (the patients with moderately health problems who could benefit considerably), compared to the 21% reported by Ubel, and between 19.7% and 25% of respondents (22% when pooled) preferred to direct resources to group A (the patients with severe health problems who could benefit a little) compared to 6% reported by Ubel. Overall Ubel reports 79% preferring either the severely affected group or to divide resources equally between the two groups, and the current study (Q1–Q2) finds between 59–63% (60% when pooled) support this preference profile, regardless of the limited help available from treatment.

Table 3 Sample characteristics (percentage) by question group (Q1–Q4).

Total (number) Gender Male Female Age 18–34 35–54 55þ Social grade AB C1 C2 DE Work Status Working full time Not full time Working part-time House person Retired Unemployed Permanently sick/disabled Student Household income Below £17,500 £17,500–29,999 £30,000–49,999 Above £50,000 Refused/not stated Ethnicity White British BMEa/other Home ownership Owner occupier Social renter Private renter Illness/disability Yesb No Household composition With children Without children Health in general Good or very good Fair Bad or very bad Health Insurance Yes

Question Q1

Question Q2

Question Q3

Question Q4

68

66

63

64

44.1 55.9

43.9 56.1

49.2 50.8

42.2 57.8

29.4 39.7 30.9

28.8 34.8 36.4

41.3 25.4 33.3

40.6 37.5 21.9

14.7 35.3 19.1 27.9

13.6 25.8 27.3 31.8

15.9 22.2 25.4 31.7

10.9 23.4 26.6 39.1

39.7 60.3 11.8 10.3 23.5 1.5 5.9 7.4

33.3 66.7 13.6 6.1 34.8 6.1 0.0 6.1

36.5 63.5 15.9 9.5 20.6 4.8 6.3 6.3

34.4 65.6 23.4 7.8 12.5 1.6 9.4 10.9

23.5 20.6 17.6 10.3 27.9

37.9 21.2 13.6 1.5 25.8

41.3 15.9 14.3 4.8 23.8

37.5 14.1 14.1 25.0 26.6

80.9 19.1

89.4 10.6

82.5 17.5

82.8 17.2

63.2 20.6 14.7

53.0 25.8 18.2

54.0 23.8 20.6

56.3 21.9 18.8

39.7 54.4

48.5 50.0

38.1 60.3

31.3 65.6

30.9 67.6

28.8 68.2

27.0 69.8

43.8 56.3

73.5 14.7 10.3

74.2 15.2 9.1

66.7 20.6 9.5

70.3 21.9 7.8

14.7

10.6

14.3

18.8

Chi-sq 0.872

0.419

0.724

0.173

0.528*

0.557

0.227

0.294

0.216

0.919

0.63

*Combining £30,000–49,000 and Above £50,000. a Black and minority ethnic. b Respondent and/or someone else in household.

2252

C. Green / Social Science & Medicine 68 (2009) 2247–2255

Table 4 Responses to preference questionnaire (Q1–Q4). Number (percentage) of respondents expressing preference by group Version

Description

Q1

Ubel textd3 response categories; Severe health problems where treatment helps a little vs. moderate health problems where treatment helps considerably Ubel textd4 response categories; Severe health problems, treatment helps a little vs. moderate health problems where treatment helps considerably Adapted textd3 response categories; Disadvantaged patient group where treatment helps a little vs. more advantaged patient group where treatment helps considerably Adapted textd4 response categories; Disadvantaged patient group where treatment helps a little vs. more advantaged patient group where treatment helps considerably Total

Q2

Q3

Q4

Q1þQ2 Q3þQ4

Results combined for questions on patients with severe vs. moderate health problems Results combined for questions on disadvantaged vs. more advantaged patients

N

Group A

Group B

Equal for A & B

Others to choose

Other/Don’t know

p Valuea

(severe/disadv.)

(moderate/adv.)

66

13 (19.7)

19 (28.8)

29 (43.9)

NA

5 (76)

0.289

68

17 (25)

21 (30.9)

23 (33.8)

5 (7.4)

2 (3)

0.516

63

29 (46)

11 (17.5)

21 (33.3)

NA

2 (3.2)

0.004

64

18 (28.1)

8 (12.5)

35 (54.7)

2 (3.1)

1 (1.6)

0.050

261

77 (29.5)

59 (22.61)

108 (41.38)

10 (3.83)

0.123

134

30 (22.4)

40 (29.9)

52 (38.8)

5 (3.7)

7 (5.2)

0.232

127

47 (37)

19 (15)

56(44.1)

2 (1.6)

3 (2.4)

0.001

7 (2.68)

NA, not applicable/not contained in that version of the question. a For columns titled ‘Group A’ and ‘Group B’ only (not relevant for those with ‘Equal preferences’ or ‘Other’: Chi-squared test of proportions preferring A compared to those preferring B).

However, results are generally consistent with those presented by Ubel, in so far as the clarification (amendment) to the text used by Nord (1993a) resulted in a significant difference in the percentage choosing group B, in both Ubel (1999) and in this study i.e. 21% and 30% respectively, compared to less than 10% in studies by Nord (1993a) and Ubel (replication of the Nord question). In this survey, where the groups have been labelled differently, to present the choice as one between patients who are disadvantaged with treatment helping a little, versus patients who are more advantaged with treatment helping considerably (Q3 and Q4), the response data show a preference to treat the disadvantaged [‘worse-off’] group. In these questions, between 28% and 46% of respondents indicated that they preferred to treat the disadvantaged group, and between 33% and 55% preferred to divide the resources equally between the two groups. Only 12.5–17.5% preferred to direct resources to the more advantaged patients, regardless of their greater benefit from treatment. Considering the combined results (Table 4), combined by group description, there is a stronger preference to support the worse off group when it is described using terminology of ‘disadvantaged’ versus ‘more advantaged’ (Q3 and Q4), compared to the severity based questions. Combining responses from Q3 and Q4, 37% expressed a preference to treat the disadvantaged patient group, with only 15% preferring the more advantaged patient group, who could get greater help (benefit). Over 80% of respondents (Q3 and Q4) preferred to give at least equal preference to the two groups, even though patient group A were only able to benefit ‘a little’ from treatment. The preference to treat the worse off group, who could only benefit ‘a little’ (compared to considerable benefit) was weaker when the group were described according to severity of health condition (Q1 and Q2). When combining response data from all four of the subgroups, regardless of descriptions used and response categories available, there is a strong preference to give group A (the worse-off group) at least equal priority, with over 70% of respondents selecting either A or the equal priority option. Although, a substantial proportion (over 20%) of respondents show a preference to treat those who are better off and able to get the greater help from treatment.

Discussion This study has elicited the preferences of a sample of the UK general public over healthcare priority setting scenarios. It used a random representative sample, and face-to-face interviews, and it extends the methodology previously reported by Ubel (1999) through the use of alternative descriptions for groups competing for the same healthcare resources. The study adds to the empirical ethics literature, providing insights on the preferences of the public over the use of severity of health as a basis for setting priorities, and against the meaning of a severity of health criterion in the broader sense of distributive preference and fairness, in resource allocation. Findings add to the evidence base that indicates society (general public) does not support a strict health maximisation (efficiency) objective when it comes to difficult priority setting choices, and the allocation of health care resources. The findings from this study offer support for the severity hypothesis (Nord, 2005). The results show that where a patient group with severe health problems is not able to benefit greatly from treatment, compared to a group competing for the same resources that have moderate health problems and that are able to show greater benefits from treatment, the majority of respondents wish to give at least equal priority to the more severely affected patient group. Such preferences suggest that a unit of health gain for the more severely affected group (Group A) is given greater social value than the same unit of health gain for the less severely affected group (group B), which is the premise of the severity hypothesis. Using a different label for groups competing for resources, the study has explored a more general hypothesis that respondents will prefer to give priority to those groups that may be regarded as the worse-off groups, regardless of the basis for being worse-off. Respondents have given a strong indication that worse-off groups, regardless of specific arguments surrounding severity of health or disadvantage, should; (i) at least be treated equally, even though they are able to get only a little help (compared to considerable) from treatment, (ii) could be considered for greater priority, with 22% to 37% of respondents preferring to give priority to the

C. Green / Social Science & Medicine 68 (2009) 2247–2255

worse-off groups, compared to between 15% and 30% who preferred to give priority to the better-off groups. This latter finding should be interpreted against the differential help available, to alternative groups, from available treatment. Results indicate that respondents value health gain in (and treatment of) the worse-off groups more highly than health gain in (and treatment of) those in the less severe and more advantaged groups. In earlier studies by Nord (1993a) and Ubel (1999), particularly this latter study, and in a number of other empirical studies around distributive preference (Dolan et al., 2005), the majority of respondents indicate an equal preference against two competing scenarios presented in the priority-setting questions. Given the absence of alternative response categories in these studies, the general meaning of this common ‘equal’ response is uncertain. One suggestion is that a stated equal preference may have been used to avoid a difficult decision between two competing options. In this study, very few respondents opted to avoid the difficult decision by stating the ‘choice should be made by others’. Given the findings presented here, it seems more likely that a preference to give equal priority can be interpreted as a true preference for equality. The finding may be generalisable, across other areas of the empirical ethics literature, although it remains experimental and indicative, and should be explored more widely in future studies. Whilst Nord has championed the use of severity of health as a basis for setting priorities, he has also supported the notion that there is a broader concern for the worse-off (Nord, 2005). There is general moral support for such a view, with, for example, Rawls (1971) often cited as suggesting that a theory of justice should place special importance on the position of the worse off in society. Parfit (1991) and Brock (1991) have offered a rationale for giving greater concern to the worse-off groups, suggesting they may have more urgent needs. Whilst the evidence around the severity hypothesis may be an accurate reflection of social values, it may also be the case that it is not severity of health per se that is driving the preferences given. It may be a broader preference to support those regarded as worstoff, and such a social value would draw together much of the empirical literature around health care priority setting, to reflect a general preference for ‘fairness’. Such a preference for fairness is evident in the broad literature that argues against the use of efficiency arguments, in the form of health gain maximisation, in heath care priority setting decisions (Dolan et al., 2005; Green, 2007). The literature does also indicate potential thresholds around where and when it may be appropriate to give priority to the better off groups (e.g. where benefits from treatment are not regarded as meaningful to patient groups). The current study, offers some support for a more general preference to treat the worse-off groups, regardless of the way these groups are described. The findings here indicate that when patient groups are described in a more general way (not limited to health) there may be a stronger preference towards the worse-off groups. Findings here support a potential preference for ‘fairness’ in the treatment of worse-off groups, but the study does not explain what is meant by fairness (or equity), and it is still down to empirical testing to determine what balance of gains and losses is an acceptable presentation of fairness. Given the potential influence of the current empirical ethics literature in the setting of health policy, exploring the literature more fully and developing a greater understanding around public preferences is important. Some commentators are suggesting that the empirical ethics literature should be used to inform health policy in a more direct way (Richardson & McKie, 2005), and organisations such as the National Institute of Health and Clinical Excellence (NICE) in the UK, are looking at the empirical ethics literature for guidance on criteria that may be important when setting out social value judgments to inform decisions (e.g. NICE, 2005, 2008a). In the UK, the

2253

NICE Citizens Council has recently considered the issue of severity of health, and whether NICE should take it into account when making its decisions (NICE, 2008b). Other policy forums will, likewise, face similar challenges, therefore, considering the empirical evidence around severity of health, and examining the broader meaning of preferences stated in favour of severe versus less severe health problems is particularly relevant. The suggestion here that studies reporting a distributive preference in favour of severe health problems may be representing a broader preference for fairness in favour of worse-off groups, and that they are not necessarily about severity of health per se, highlights that some caution should be exercised when interpreting the current empirical ethics literature. Not that severity of health is unimportant, but that preferences for severity of health can be interpreted in a number of ways. Nord (2005) has discussed the notion of worse-off, and the fact that if priority is to be given to the worse-off, it may be a question of whether we should be concerned for those worse-off in terms of health or those worse-off in some other life situation. Nord suggests, presumably informed in some way by the empirical ethics literature, that there is greater agreement about giving priority to those who are worst-off in terms of health. However, the findings in the current study offer empirical evidence to challenge that view. That severity of health may reflect some form of natural inequality, related to health status, rather than a social inequality related to social circumstances, as represented by ‘disadvantage’ (e.g. low income family) in the current study, is worthy of further investigation (see Rawls, 1971; Kymlicka, 2001, for some general discussion). Whether some person is in need of health care (and in a way worse-off) as a consequence of their own actions is an issue that has been considered in the empirical ethics literature (e.g. Olsen, Richardson, Dolan, & Menzel, 2003), and it is an issue that has been considered by policy makers (NICE, 2005, 2008a). However, more generally, it is becoming evident that the current literature on empirical ethics is becoming more important and influential in a health policy setting, as policy makers grapple with the challenges of priority-setting in an increasingly open and transparent way. It is therefore going to be important for policy makers to consider the literature more fully, across a greater number of potential priority setting criteria, both health-related and non-health-related. It will be necessary to identify the gaps in the literature, as well as the specific contributions available, and to further interpret the meaning of the current evidence around empirical ethics. The current study has limitations. The study uses hypothetical scenarios, although this is difficult to avoid given the nature of priority setting choices in a sample of the general public. The current study uses previous studies, and study frames, presented by Nord (1993a) and Ubel (1999), and is therefore limited in the text presented and subject to the framing and context effects present in the original text. As with earlier studies from Nord and Ubel, it is difficult to quantify the differences between alternative treatment scenarios, with differences represented by qualitative labels (‘a little’ versus ‘considerable’ help). Earlier studies (those of Nord and Ubel) have regarded the differences presented as ‘extreme’ scenarios, although this is down to subjective judgment and interpretation (and deserving of further qualitative research and investigation). As with earlier studies, the current study uses simple scenarios, and asks respondents to indicate a preference using the general allocation of a hypothetical budget (constrained). Given the simple experimental methods used, Nord (1993a, 1999) and Ubel (1999) have stated that the findings from their studies have no immediate policy implications, as the text used is judged to be ‘too vague’ to guide policy decisions (e.g. Ubel, 1999, p. 901). However, these studies have been influential in the health care literature, and potentially in health policy forums acting on the available

2254

C. Green / Social Science & Medicine 68 (2009) 2247–2255

literature. The current study has used a more formal and rigorous sampling frame and interview techniques, and the questions were asked at the end of a 20-minute health-related priority setting interview. During the interview respondents had had the opportunity to consider priority setting scenarios within the UK NHS, and to deliberate on the relative merits of ‘severity of health condition’ as a priority setting criterion, as well as other criteria (including efficiency and non-health attributes). This is thought to be the first such study to consider together the issues of severity of health condition and disadvantage, in the context of health care resource allocation and priority-setting. The current study has used the different descriptions of healthcare groups to investigate the notion of ‘worse-offness’ more broadly, but the study did not consider how respondents would trade-off between these two competing groups, described using severity arguments versus disadvantage. Given the results presented here, showing the feasibility and acceptability of this type of survey, using different descriptions for health and non-health, and that such opposing views are likely to come into conflict in a health policy setting, further research to quantify potential trade-offs is recommended. Conclusions This study presents evidence to support the importance of severity of health when allocating health care resources, and the setting of health care priorities. It also provides empirical insights into the interpretation of a commonly reported stated preference, in studies around distributive preferences, to allocate resources equally across competing groups. Findings presented suggest that such a preference may indeed be a true preference for equality rather than respondents avoiding difficult decisions. This study also suggests some support for a view that the empirical ethics literature around severity of health should be interpreted more broadly. Findings indicate that a public preference to treat those groups most severely affected by their health may be reflecting a more general preference towards fairness, in support of the needs of those groups classed as the worse-off groups. Acknowledgements I would like to thank the Ipsos MORI Social Research Institute for undertaking data collection, especially Gregor Jackson. I acknowledge financial support from the Wessex Institute for Health Research and Development, University of Southampton, for funding to undertake data collection. Early versions of this paper have benefited from comments from Dr Karen Gerard, and Professor James Raftery, University of Southampton, and special thanks for helpful comments from Dr Erik Nord, Norwegian Institute of Public Health, and two anonymous referees. Appendix A. Example of showcard used in interviews/survey

ILLNESS A ILLNESS B Disadvantaged patient group More advantaged patient group (e.g. from a low income family) Treatment helps patients Treatment helps patients a little considerably Both patient groups have similar health problems without treatment. Cost of treatment is the same in both cases. Unable to treat all patients in both groups.

(1) Most of the increase should be allocated to treatment for illness B, in a more advantaged patient group, which improves patients considerably with treatment

(2) Most of the increase should be allocated to treatment for illness A, in a disadvantaged patient group, which improves patients a little with treatment (3) The increase should be divided evenly between the two groups (4) I am not able to make a decision and would prefer that the choice be made by others

References Brock, D. (1991). Priority to the worst off in health care resource prioritisation. In M. Battin, R. Rhodes, & A. Silvers (Eds.), Health care and social justice. New York: Oxford University Press. 1991. Cookson, R., & Dolan, P. (1999). Public views on health care rationing: a group discussion study. Health Policy, 49(1–2), 63–74. Culyer, A. J. (2001). EquitydSome theory and its policy implications. Journal of Medical Ethics, 27(4), 275–283. Dolan, P. (1998). The measurement of individual utility and social welfare. Journal of Health Economics, 17(1), 39–52. Dolan, P., Shaw, R., Tsuchiya, A., & Williams, A. (2005). QALY maximisation and people’s preferences: A methodological review of the literature. Health Economics, 14(2), 197–208. Green, C. (2007). Justice, fairness and equity in health care: exploring the social value of health care interventions. PhD Thesis, University of Southampton. Gyrd-Hansen, D. (2004). Investigating the social value of health changes. Journal of Health Economics, 23(6), 1101–1116. Hauck, K., Smith, P. C., & Goddard, M. (2003). The economics of priority setting for health care. A Literature review. Washington, USA: Health Nutrition and Population (NHP) Discussion Paper, The World Bank. Kymlicka, W. (2001). Contemporary political philosophy: An Introduction (Second edition.). Oxford: Clarendon Press. Newdick, C. (2005). Who should we treat? Rights, rationing, and resources in the NHS (2nd Edition). Oxford, New York: Oxford University Press. NICE (UK National Institute for Health and Clinical Excellence). (2005). Social value judgements. Principles for the development of NICE guidance. London: National Institute for Health and Clinical Excellence. www.nice.org.uk. NICE (UK National Institute for Health and Clinical Excellence). (2008a). Social value judgements. Principles for the development of NICE guidance (2nd Edition). London: National Institute for Health and Clinical Excellence. www.nice.org.uk. NICE(UK National Institute for Health and Clinical Excellence). (2008b). Report on NICE Citizens Council meeting: quality adjusted life years (QALYs) and the severity of illness. London: National Institute for Health and Clinical Excellence. www. nice.org.uk. Nord, E. (1993a). Norwegian health care politicians do not wish to maximize health. Journal of the Norwegian Medical Association, 113, 1371–1373, (in Norwegian). Nord, E. (1993b). The trade-off between severity of illness and treatment effect in cost- value analysis of health care. Health Policy, 24(3), 227–238. Nord, E. (1999). Cost-value analysis in health care: making sense out of QALYs. Cambridge: Cambridge University Press. Nord, E. (2005). Concerns for the worse off: Fair innings versus severity. Social Science and Medicine, 60(2), 257–263. Nord, E., Richardson, J., Street, A., Kuhse, H., & Singer, P. (1995). Maximizing health benefits vs egalitarianism: An Australian survey of health issues. Social Science and Medicine, 41(10), 1429–1437. Oddsson, K. (2003). Assessing attitude towards prioritizing in healthcare in Iceland. Health Policy, 66(2), 135–146. Olsen, J. A., Richardson, J., Dolan, P., & Menzel, P. (2003). The moral relevance of personal characteristics in setting health care priorities. Social Science and Medicine, 57, 1163–1172. PBAC (Australian Pharmaceutical Benefits Advisory Committee) (2004). Department of Health and Ageing. Other relevant factors and the ‘rule of rescue’; ratified at the March 2003 meeting of the Pharmaceutical Benefits Advisory Committee; 2004, Commonwealth of Australia, Canberaa. 2004 [Available from URL: http://www.health.gov.au/internet/main/publishing.nsf/Content/healthpbs-general-listing-committee2.htm, accessed Feb 2008]. Parfit, D. (1991). Equality or priority. The Lindley Lecture. Department of Philosophy, University of Kansas. Rawlins, M. D., & Culyer, A. J. (2004). National Institute for Clinical Excellence and its value judgements. British Medical Journal, 329, 224–226. Rawls, J. (1971). A theory of justice. New York: Oxford University Press. Richardson, J. (2002). Empirical ethics. In C. J. L. Murray (Ed.), Summary measures of population health: Concepts, ethics, measurement and applications. Geneva: World Health Organisation. Richardson, J., & McKie, J. (2005). Empiricism, ethics and orthodox economic theory: what is the appropriate basis for decision-making in the health sector? Social Science and Medicine, 60, 265–275. Sassi, F., Le Grande, J., & Archard, L. (2001). Equity versus efficiency: A dilemma for the NHS. British Medical Journal, 323(7316), 762–763. Sassi, F., Archard, L., & Le Grande, J. (2001). Equity and the economic evaluation of healthcare. Health Technology Assessment, 5(3).

C. Green / Social Science & Medicine 68 (2009) 2247–2255 Sen, A. (2002). Why health equity? Health Economics, 11, 659–666. Schwappach, D. L. (2002). Resource allocation, social values and the QALY: a review of the debate and empirical evidence. Health Expectations, 5(3), 210–222. Southampton City Council. (2004). The 2001 Census Results. Southampton City Council (July), www.southampton.gov.uk.

2255

Ubel, P. A. (1999). How stable are people’s preferences for giving priority to severely ill patients? Social Science and Medicine, 49(7), 895–903. UK Office for National Statistics. (2002). Disadvantaged households. Results from the 2000 General Household Survey. London: National Statistics. UK Office for National Statistics. (2004). Focus on social inequalities. London: National Statistics.