2. Valuing health at the end of life
22/09/15 2
• This research is a collaboration between Koonal Shah
(Office of Health Economics; University of Sheffield) and
Professors Aki Tsuchiya and Allan Wailoo (both University
of Sheffield)
• The literature review reported here is in-progress and its
results should be treated as preliminary
• The views, and any errors or omissions, expressed are of
the presenting author only
Study status and acknowledgements
3. Valuing health at the end of life
22/09/15 3
Criteria that need to be satisfied for NICE’s supplementary
end of life policy to apply are currently as follows:
NICE end of life criteria
C2
The treatment is indicated for patients with a short
life expectancy, normally less than 24 months
There is sufficient evidence to indicate that the
treatment offers an extension to life, normally of at
least an additional three months, compared to current
NHS treatment
The treatment is licensed or otherwise indicated, for
small patient populations
C3
C1
4. Valuing health at the end of life
22/09/15 4
• Placing additional weight on survival benefits in patients
with short remaining life expectancy could be considered a
valid representation of society's preferences
• But the NICE consultation revealed concerns that there is
little scientific evidence to support this premise
• Two (unpublished) reviews of the stated preference /
empirical ethics literature undertaken in 2011 did not
identify many relevant studies
NICE end of life criteria (2)
5. Valuing health at the end of life
22/09/15 5
• To review the published literature that is relevant to the
following research question:
Do members of the general public wish to place
greater weight on a unit of health gain for end of life
patients than on that for other types of patients?
• To identify the extent to which public preferences on this
topic have been studied in the peer-reviewed literature
• To provide an in-depth account of the methods used to
elicit preferences and the findings of the studies, with the
intention of informing policy decisions and future research
in this area
Objectives
6. Valuing health at the end of life
22/09/15 6
• Primary source of data: electronic search of the Social
Sciences Citation Index (SSCI) within Web of Science
• Follow-up of reference lists of articles identified using the
final SSCI search
• Articles already known to me that met the criteria for
inclusion
Data sources
7. Valuing health at the end of life
22/09/15 7
• Search terms developed using an iterative process
• Base search terms: end of life AND preferences – 1,076
results
• Added two alternative terms related to end of life:
severity and terminal – 2,387 results
• Designation of three studies already known to me as ‘key
papers’ – examined abstracts of the key papers to identify
further search terms (see next slide)
Search terms
9. Valuing health at the end of life
22/09/15 9
• Search terms developed using an iterative process
• Added health, respondents (+ subjects, participants,
sample), life expectancy
• Conducted informal assessments of whether the
additional records identified by adding individual search
terms contained at least some potentially relevant records
(this justifying their inclusion)
• Checked whether any key records were missed after
adding certain terms – this ruled out patient, treatment,
evidence, public and population
• Also checked random sample of additional records
identified by adding synonyms for preferences (attitudes,
choices, utilities, values)
Search terms (3)
10. Valuing health at the end of life
22/09/15 10
("end of life" OR severity OR terminal OR “life expectancy”)
AND preferences
AND health
AND (respondents OR subjects OR participants OR sampl*)
• Yielded 598 unique results
Final strategy
11. Valuing health at the end of life
22/09/15 11
To be included, articles had to meet all of the following sequential criteria:
1. Publication: Article must be published in English in a peer-reviewed
source.
2. Empirical data: Article must review, present or analyse empirical data.
3. Priority-setting context: Article must relate to a health care priority-
setting or resource allocation context.
4. Stated preference data: Article must report preferences that were
elicited in a hypothetical, stated context using a choice-based approach
involving trade-offs.
5. End of life: Article must address the topic of giving priority to end of life
patients (i.e. patients with short life expectancy) or to treatments for
such patients.
6. Original research: Article must present original research and must not
be solely a review of the literature.
Selection of studies for inclusion
12. Valuing health at the end of life
22/09/15 12
To be included, articles had to meet all of the following sequential criteria:
1. Publication: Article must be published in English in a peer-reviewed
source.
2. Empirical data: Article must review, present or analyse empirical data.
3. Priority-setting context: Article must relate to a health care priority-
setting or resource allocation context.
4. Stated preference data: Article must report preferences that were
elicited in a hypothetical, stated context using a choice-based approach
involving trade-offs.
5. End of life: Article must address the topic of giving priority to end of life
patients (i.e. patients with short life expectancy) or to treatments for
such patients.
6. Original research: Article must present original research and must not
be solely a review of the literature.
Selection of studies for inclusion
13. Valuing health at the end of life
22/09/15 13
• Author(s)
• Year of publication
• Country of origin of data
• Sample size
• Type of sample
• Sample recruitment process
• Exclusion criteria
• Mode of administration
• Objective(s)
• Pilot reporting
• Preference elicitation technique
• Perspective
• End of life definition
• Life expectancy without treatment
attribute levels
• Life expectancy gain from
treatment attribute levels
• Was disease labelled/named?
• Did the study examine whether
quality of life improving or life
extending treatments are
preferred?
Data extraction fields
14. Valuing health at the end of life
22/09/15 14
• What were respondents choosing
between (or choosing to do)?
• Were indifference options offered?
• Were visual aids used?
• Was strength of preference
examined at the individual level?
• Number of tasks completed by
each respondent
• Time taken to complete survey
reported?
• Summary of finding: end of life
vs. non-end of life
• Summary of finding: quality of life
improvement vs. life extension
• Other results of potential interest
• Other factors examined
• Impact of background
characteristics
• Were qualitative data or
explanatory factors sought?
• Was any reference made to age-
related or time-related
preferences?
Data extraction fields (2)
16. Valuing health at the end of life
22/09/15 16
Summary of included studies (n=17)
Authors (date) Country Sample size (type) Method Mode Summary of primary study objective(s)
Abel Olsen (2013) NOR 503 (public) Choice Internet survey To test for support for end of life prioritisation and the fair innings approach
Baker et al. (2010) UK 587 (public) DCE CAPI To test for support for multiple prioritisation criteria
Dolan and Cookson
(2000)
UK 60 (public) Choice Focus groups Qualitative examination of support for multiple prioritisation criteria
Dolan and Shaw
(2004)
UK 23 (public) Choice Focus group To test for support for multiple prioritisation criteria
Dolan and Tsuchiya
(2005)
UK 100 (public) Choice;
ranking
Self-completion survey To compare support for prioritisation according to age vs. prioritisation
according to severity/life expectancy
Kvamme et al. (2010) NOR 2,143 (public) WTP Internet survey To test for non-linear utility of short life extensions from an individual
perspective
Lim et al. (2012) KOR 800 (public) DCE Internet survey To test for support for multiple prioritisation criteria
Linley and Hughes
(2013)
UK 4,118 (public) Budget
allocation
Internet survey To test for support for multiple prioritisation criteria
Pennington et al.
(2015)
Multiple 17,657 (public) WTP Internet survey To compare WTP for different types of QALY gain
Pinto-Prades et al.
(2014)
SPA 813 (public) WTP;
PTO
CAPI To test for support for end of life prioritisation and to compare support for
life extensions vs. quality of life improvements
Richardson et al.
(2012)
AUS 544 (public) Other Internet survey and
self-completion survey
To test a technique for measuring support for health-maximisation and
health sharing
Rowen et al. (2015) UK 3,669 (public) DCE Internet survey To test for support for multiple prioritisation criteria
Shah et al. (2014) UK 50 (public) Choice Person interview To test for support for end of life prioritisation
Shah et al. (2015) UK 3,969 (public) DCE Internet survey To test for support for end of life prioritisation
Skedgel et al. (2015) CAN 656 (public, decision-
makers)
DCE Internet survey To test for support for multiple prioritisation criteria
Stahl et al. (2008) USA 623 (public) Choice Internet survey To test for support for multiple prioritisation criteria
Stolk et al. (2005) NLD 65 (students,
researchers, health
policy makers)
Choice Personal interview To test for support for multiple approaches to priority-setting
17. Valuing health at the end of life
22/09/15 17
Summary of findings
Freq. %
Overall finding: end of life premium
- Evidence consistent with an end of life premium
- Evidence not consistent with an end of life premium
- Mixed or inconclusive evidence
7
7
3
41.2%
41.2%
17.6%
Overall finding: QOL-improving vs. life-extending end of life treatments
- QOL improvement preferred
- Life extension preferred
- Not examined / reported
2
1
14
11.8%
5.9%
82.4%
18. Valuing health at the end of life
22/09/15 18
Distribution of selected variables, by
overall study finding
* Study combining PTO and WTP methods counted as two studies since separate results are reported for both. Study combining
ranking exercise and other choice exercise counted as one study since this is considered to be a single hybrid method.
Variable Evidence consistent with
an end of life premium
Evidence not consistent
with an end of life
premium
Country
- UK
- Europe (non-UK)
- Rest of the world
2
3
2
4
2
1
Method*
- DCE
- Other choice exercise
- Willingness to pay
- Other
2
2
3
1
2
3
0
2
Mode of administration
- Internet survey
- Other
5
2
4
3
Indifference option(s) offered?
- Yes
- No or not reported
5
2
1
6
Visual aids used?
- Yes
- No or not reported
5
2
2
5
19. Valuing health at the end of life
22/09/15 19
Discussion point: choice of method
• Most studies asked respondents to adopt a social decision maker
perspective and to answer questions typically of concern to a
health care decision maker
• Respondents are expected to answer the questions based on
what they consider to be appropriate and acceptable for society
– would not necessarily expect to benefit personally from their
choices
• The three WTP studies asked respondents to adopt an individual
or ‘own health’ perspective
• Consistent with the welfarist view that confines the evaluative
space to individual utility only
• But are WTP valuations made by individuals facing the prospect
of imminent death a useful way of guiding decisions about how
to spend a common pool of funding?
20. Valuing health at the end of life
22/09/15 20
Discussion points: indifference
options and visual aids
• Studies offering opportunity to express indifference between alternatives
were more likely to report evidence consistent with an end of life
premium than those that did not
• Way in which indifference options are framed may affect respondents’
willingness to choose those options
• e.g. 50:50 split of resources vs. ‘I have no preference’
• When no indifference option is available, a respondent who is indifferent
may pursue a strategy other than making choices at random
• Trend towards DCEs administered via internet surveys suggests that
indifference options may become less frequent
• Studies that used visual aids were more likely to report evidence
consistent with an end of life premium that those that did not
• Could graphical representations unintentionally lead to different
respondents interpreting the information in different ways?
22. Valuing health at the end of life
22/09/15 22
Discussion points: age and time-
related preferences
• Majority of studies included patient age in the study design
• In some cases age was one of several prioritisation criteria being examined; in
other cases researchers sought to examine whether respondents’ end of life-related
preferences were influenced by the ages of the patients
• Some evidence that respondents become less concerned about the number of life
years remaining when the patients in question are relatively old
• UK policy context: age cannot be used as a priority-setting criterion
• Few studies mentioned time-related preferences; even fewer attempted to control
for them
• Could an observed preference for treating patients with short life expectancy be
driven by a preference for benefits occurring sooner rather than later?
• Could an observed preference for treating patients with short life expectancy be
driven by concern about how long those patients have to ‘prepare for death’?
23. Valuing health at the end of life
22/09/15 23
Both patients are same age today (Time=0)
Age
denotes time in full quality of life
denotes life extension (at full quality of life) achievable from treatment
Time
(years) 0 1 2 3 4 5 6 7 8 9 10 11
Patient A
Patient B
Source: Shah et al. (2014)
24. Valuing health at the end of life
22/09/15 24
Patient B is 9 years older than patient A today
Age (2)
Time
(years) 0 1 2 3 4 5 6 7 8 9 10 11
Patient A
Patient B
Source: Shah et al. (2014)
25. Valuing health at the end of life
22/09/15 25
Both patients are same age today
Time
Time
(years) -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2
A
B
Source: Shah et al. (2014)
26. Valuing health at the end of life
22/09/15 26
Limitations of the review
• Searched only one database – covers all major health economics and
health policy journals but may not cover specialist scientific journals
• Included only articles that have been published in English, and the
review was (to a large extent) motivated by the policy context in the UK
• Search terms considered are likely to reflect the language used by health
economics researchers in the UK – may not be well suited for identifying,
say, articles authored by ethicists or researchers based in LMICs
• No formal assessment of study quality – used publication in a peer-
reviewed source as a proxy for quality
• Assignment of study findings into categories (consistent; not consistent;
inconclusive) involved a degree of subjective judgement
• No clear definition of what counts as “consistent with an end of life
premium”
• Unanimous preferences are rarely observed in stated preference studies
• Not always possible to use authors’ own conclusions as a guide
27. Valuing health at the end of life
22/09/15 27
Conclusions
• Primary finding of the review is that the existing evidence is mixed
• Reviews of severity-related preferences more generally have been able
to make more decisive conclusions – i.e. an overall preference for giving
priority to treating those who are terminally ill
• Suggests that people are more concerned about treating patients with
poor quality of life than treating patients with short life expectancy – but
this supposition is not supported by individual studies that examined both
simultaneously
• Gaps in the literature / recommendations for further research
• Given the known issues associated with framing effects, researchers could
use multiple methods or designs to test the robustness of their results
• It would be informative to seek to understand the extent to which
respondents agree with researchers’ interpretations of their choices
• Preferences regarding ‘preparedness’ have received limited attention in
the literature to date – further investigation would be welcomed