Aileen Clarke and Sian Taylor-Phillips' presentation development of a preference based well-being measure for the CLAHRC WM Scientific Advisory Group, 10th June 2015, Birmingham, UK
2. Prevention and Detection
Theme 3 – CLAHRC West Midlands
Division of Health Sciences
Warwick Medical School
The research was funded by the National Institute for Health Research (NIHR)
Collaboration for Leadership in Applied Health Research and Care (CLAHRC) West Midlands.
3. “Where’s WALY” team
Jason Madan
Rebecca Johnson
David Jenkinson
Sian Taylor-Phillips
Wendy Robertson
Hendramoorthy Maheswaran
Stavros Petrou
Sarah Stewart-Brown
Aileen Clarke
Advisory group: Paul Dolan (LSE), Ewen Mackinnon (Well-being and
Civil Society Policy Team, Cabinet Office), Paul Litchfield (Chair, What
Works - Centre for Wellbeing)
4. Well-being as a cross-sectoral measure
of benefit.
Instruments such as the EQ-5D allow policy-makers to
prioritise health spending based on public values.
The use of EQ-5D – derived QALYs to maximise the benefits
from health spending is well established in health
technology assessment
Many public sector interventions improve lives in ways that
are not adequately captured by EQ-5D
Well-being is a concept that captures many of the ways in
which public sector spending benefits populations (health,
safety, social connections, etc.)
The Warwick-Edinburgh Mental Wellbeing Scale
(WEMWBS) has been widely used to measure well-being.
5. The problem
There are outcomes and effects of interventions other
than on health
Decisions may impact on other sectors: wider social
benefits/effects
– social networks, relations, friends, family income
– public health, social care, transport, leisure, agriculture
Even in health care, EQ-5D has been questioned. May not
capture all the aspects of quality of life that matter……
So….
6. Where’s WALY – development of a
preference based well-being measure
Research is planned to develop a preference-based tariff
for the WEMWBS.
This will allow calculation of WALYs for economic
evaluations.
The tariff will be based on responses from a large
representative UK sample.
Follow-on research questions include:
– How do values vary across populations (e.g. service users, general
population)?
– What types of intervention give benefits that should be measured
in terms of WALYs?
7. Study design for estimating the
WALY tariff
Tariff will be based on the short form - sWEMWBS to
reduce the number of states (78125 vs 6.1 billion)
Subset selected for valuation (balanced on overall
severity and mix of severity across dimensions).
Hybrid TTO – DCS method will be used for preference
elicitation
Use professional survey organisation to ensure
representative sample.
Minimum of 25 participants per health state (so 200
health states would imply sample size of 5000+)
8. Pilot work:
– To investigate relationship between EQ-5D and
WEMWBS: capturing differences and capturing
change:
• In Coventry Household Survey and Health Survey
for England – overall
• in subsets (general health, healthy behaviours,
economic prosperity, community, satisfaction with
neighbourhood)
• In 2 datasets where change has occurred
9. Data
Data from Coventry Household Survey
(CHS)
Year 2011 2012 2013 Total
Cases 3144 2117 2208 7469
No individual has taken part in more than one
year
11. WEMWBS and EQ-5D
Warwick Edinburgh Mental Well-being
Scale - WEMWBS
– 14 items, 5 point scale for each item
– Sum scores across the 14 items; total - 14-70
EQ-5D: total score: 0-1
– five dimensions, scored between 1-3 and a VAS
– mobility,
– self-care,
– usual activities,
– pain or discomfort and
– anxiety or depression
16. Mapping
Linear Model.
– EQ-5D as the response (dependent) variable
– WEMWBS as the covariate (independent variable)
Term Coefficient 95% CI p-value
Constant 0.468 0.438 0.498 <0.001
WEMWBS 0.008 0.008 0.009 <0.001
ANOVA Sum of Squares df Mean Square F Significance
Regression 38.86 1 38.862 846.84 <0.001
Residual 336.33 7329 0.046
Total 375.19 7330 R2 0.104
17. Capturing Differences on CHS
General health
– Single scale rating
Healthy Behaviours
– 5 or more fruit & veg/day and
– 3 or more exercise/weeks and
– Never smoked
Economic prosperity
– Own home and
– Degree level qualification and
– F/T employment
Community
– Feels very safe during day and at night and
– Very satisfied with neighbourhood as a place to live
23. Conclusions
WEMWBS explains little of the variation of EQ-5D.
In one between-subjects dataset we found
– WEMWBS may be better than EQ5D at detecting very good self
rated health, because EQ5D has ceiling effects
– WEMWBS was superior in measuring differences in healthy
behaviours and community satisfaction between subjects
Next steps i) to determine whether WEMWBS is superior
in detecting differences within subjects before and after
a range of community, economic and health related
interventions ii) to develop the tariff
24. What does this mean…
Individuals can be in ‘good’ health but
wellbeing can be low.
Interventions can improve wellbeing
WALYs are needed to prioritise public
sector spending on such interventions.
At lower end of health scale EQ5D is better at determining self reported health states – that is what it is designed to do
At top end of health scale EQ5D had ceiling effects and wemwbs may even be better at distinguishing between health states.
Left hand side of curve says that when threshold is high wemwbs is better i.e. EQ5D ceiling effect. See straight line down as no further points. So when EQ5D score is high it is worse at distinguishing between people in very good vs other self reported health than wemwbs, despite wemwbs measuring something else.
Reminder this is 5 fruit n veg AND 3 exercise/week AND never smoked
So wemwbs is better at distinguishing between people who have very healthy behaviours from the general population
Bear in mind neither is an amazing test, but we wouldn’t expect that because they aren’t designed to directly measure this.
Here EQ5D is better, probably because ill people are not in full time employment