2. content
1. Example from the Community Wellbeing
group: Housing First (HF) evaluation.
2. Example of the Cross-cutting team: the
impact of the Improved Access to
Psychological Therapies.
3. Example from the Work and Learning
Group: problem solving training for stewards.
3. In general excel format….
Period Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Combined-discounted
total
Outcome 1
Outcome 2
Outcome 3
Outcome 4
Overal LS effect per person (LS type) A
Intended outcome label ….
Internal cost per person
External cost per person
Total cost per person B
Cost-effectiveness Pounds per unit of LS B/A
Notes: 'internal cost' means cost and negative costs born by the entity deciding on the intervention
Notes: 'external cost' means cost and negative costs born by an entity not deciding on the
intervention
4.
5. Housing First
Housing First is a form of high-
intensity support for the
homeless, renting on the private
market to get them into
permanent housing, with
intensive ongoing MH support.
Targeting in the UK has been for
high-problem cases.
6. Methodology
In order to get LS improvements, the Community Wellbeing
team looked at similar experiments in Canada, particularly
Toronto. The Stegiopoulous et al. (2015) study was used to
get estimated improvements in LS for year 1 and year 2.
In order to get at costs, the Community Wellbeing team took
as baseline the costs within the UK of housing first relative to
usual treatment by Brerethon and Piece’s (2015) analysis of 9
pilots in the UK (experienced costs, but only part of them).
For a high-level cost estimate, the Community Wellbeing team
took Blood et al. (2016) estimate of what it would cost in
Liverpool to implement the HF program. That included more
cost items than were in Brerethon and Piece (2015).
7. The Canadian experiment
2,148 homeless individuals across Canada, with new
cases starting in 2009-2011, were randomly assigned
to intensive housing-and-social-support help (versus
‘normal’) for 24 months till 2011-2013.
Each 6 months they were extensively interviewed,
with additional measures taken from public records.
Those with intensive treatments were not more likely
to stop substance abuse, had an equal or higher
number of arrests, were no better integrated in the
community, but were less likely to have emergency
department visits.
8. So the Cost-effectiveness is for…
A hypothetical Housing First Intervention in the
UK that mimicks the Canadian intervention in
terms of effects on mental health, crime, and
education. But the costs are those that would
be made in the UK.
Several items are not incorporated (like
employment). Implicitly, such effects are taken
to be 0.
9. As baseline, undiscounted, the Community
Wellbeing team estimates the cost per unit of
LS on a (0-10) scale to be 4,232 pounds.
A high estimate is 27,762 pounds per unit.
High uncertainty in all elements, meaning that
at conventional levels of thresholds, the
probability of being cost-effective is close to
20% (lower than a coin toss).
10. Period Year 2017 2018Combined-undiscounted
Intended outcome Days Stably housed 298
Other outcome 1 EQ 5D 1.51 0.08 1.59
Other outcome 2 Mental Illness Severity -0.74 0.37 0.37
Overal LS effect per person (0-6) scale 0.22 0.18 0.4
(0-10) scale 0.366666667 0.3 0.666666667
Baseline Cost Total £2,821.61
High cost estimate Total £18,508.44
Cost-Effectiveness baseline Pounds per unit of LS £4,232.42
Cost-Effectiveness High Cost Pounds per unit of LS £27,762.66
11.
12. Cross-cutting team example
We look at what a hypothetical treatment of
25% of depressed in 2010 would have meant
for Life Satisfaction, Mental Health, and costs
in the ensuing 2010-2015 period in the UK.
The intervention is based on mainly Cognitive
Behavioural Therapy intervention trials in
Doncaster, Newham, and 42-month follow
trials in the UK. This is the IAPT treatment now
being expanded to 1.5 million citizens.
13. Essential approach
Some effects can be lifted from the trials, but are only
‘known’ for specific years (loosely: year 0, 1, and 4).
Most of the effects have hence to be filled in.
The costs of the trials are ‘sort of’ known. The costs to
the public purse are based on relations estimated in
the literature and within our data.
Many (intermediate) effects are thus model-derived.
15. Contents
Model Overview
Intervention
Primary Effect
Spillover
Secondary Effects
Mediated Effects
Employment
Being
Employed
Hours
Worked
Income
Physical Health
Physical
Health
Cost Savings
Partnership
Behaviour
Public Purse
Taxes
Transfers
Health
Life Satisfaction
Mental Health
of Partner
Mental Health
of Treated
Employment
Wages
Income
Health
Market
Effects
Residual Effect
Aggre-
gation
Mediated
Effects
Reference
Effects
Final OutcomesFinal OutcomesFinal Outcomes
Model Illustration
Employment
Market OutcomesMarket OutcomesMarket Outcomes
Market OutcomesMarket OutcomesReference Points
16. What are evaluating?
A hypothetical IAPT intervention on 25% of the depressed/anxious
population of 2010. No other intervention.
The base population is Understanding Society.
GHQ12>=4 is the criterion identifying the ones in mental distress.
19. Study Intervention & Data Empirical Strategy Results
Reichman et al. (2015):
Effect of Being out of Depression
on Partnership
Employ cohort data from
Fragile Families and Child
Wellbeing Study in US
Focus on families with
children born in 20 cities
between 1998 and 2000
Baseline interview at birth,
with follow-up 1 and 3 years
(focus) afterwards
Use relationship status
(married, cohabitating)
as outcome
Use basic set of controls,
child characteristics,
prior mental health
conditions
Exploit post-partum
depression (largely random)
as exogenous variation
Post-partum depression
defined as major
depression in 12 months
following birth
Use Composite
International Diagnostic
Interview Short Form
Survey embedded into
person surveys
Stratify by previous
relationship status at
baseline
Impact of major depression
on relationship status:
− Married, if previously
married: -0.07 ppt
− Married, if previously
cohabitating: n.s.
− Married, if previously
not cohabitating: n.s.
− Not cohabitating, if
previously cohabitating:
0.11 ppt
− Cohabitating, if
previously not
cohabitating: n.s.
We take inverse impacts,
assuming equivalency
Example coefficient: effect of
Mental Health on Partnership
Note: For simplicity, so far, we assume equivalency between being married and being partnered.
Used
“Effects of maternal depression on
couple relationship status,” Review of
Economics of the Household, 13, 929-
937, 2015.
24. In terms of an excel sheet
Period Year 2010 2011 2012 2013 2014 2015Combined-discounted
Direct cost per patient
Health cost savings per patient
Welfare/tax savings per patient
Combined public purse effect per patient
Costs external to NHS per person
Treatment effect MH per patient
Treatment LS change (direct, per patient)
Partner effect MH per patient
LS change
External effect MH per patient
LS change
Employment increase patients
LS change
marriage increase patients
LS change
health increase patients
LS change
income increase patients
LS change
employment changes external
LS change
marriage change external
LS change
health increase external
LS change
income increase external
LS change
Reference income change
LS change
Reference health change
LS change
Reference employment change
LS change
All non-direct effects LS per patient
Overal LS effect per person
Cost-effectiveness Pounds per unit of LS
25. Zoom in
Period Year 2010 2011 2012
Direct cost per patient
Health cost savings per patient
Welfare/tax savings per patient
Combined public purse effect per patient
Costs external to NHS per person
Treatment effect MH per patient
Treatment LS change (direct, per patient)
Partner effect MH per patient
LS change
External effect MH per patient
LS change
26. Period Label
Direct cost per patient 650.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Health cost savings per patient 473.9241 150 192 146.6667 168.5106 150.6977
Welfare/tax savings per patient 8.2189 14.4370 6.5205 13.6176 8.9853 8.4410
Combined public purse effect per patient -167.8570 164.4370 198.5205 160.2843 177.4959 159.1387
Costs external to NHS per person
Treatment effect MH per patient -3.6296 -3.0852 -3.0852 -3.0852 -3.0852 -3.0852
Treatment LS change (direct, per patient) 0.6193 0.4967 0.4531 0.4814 0.4980 0.4823
Partner effect MH per patient -0.5444 -0.4628 -0.4628 -0.4628 -0.4628 -0.4628
LS change 0.0843 0.0600 0.0768 0.0609 0.0814 0.0674
External effect MH per patient -0.2722 -0.2314 -0.2314 -0.2314 -0.2314 -0.2314
LS change 0.0421 0.0279 0.0370 0.0292 0.0394 0.0325
Employment increase patients 2.3293 2.0883 1.6842 1.9362 2.5010 2.2316
LS change 0.0525 0.0381 0.0381 0.0442 0.0449 0.0436
single decrease patients -0.0210 -0.0165 -0.0107 -0.0106 -0.0146 -0.0127
LS change 0.0075 0.0058 0.0039 0.0037 0.0051 0.0045
health increase patients 0.9647 0.5899 0.5615 0.4990 0.4824 0.5114
LS change 0.1158 0.0661 0.0645 0.0570 0.0554 0.0588
income increase patients 88.9069 103.4365 64.0922 116.0643 130.4465 89.6381
LS change 0.0806 0.0781 0.0498 0.0680 0.0840 0.0670
Employment increase partner 0.1955 0.3798 0.1400 0.2050 0.6300 0.0639
LS change 0.0033 0.0018 0.0036 0.0009 0.0013 0.0003
single decrease partner 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
LS change 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
health increase partner 0.0944 0.0718 0.0948 0.0718 0.1059 0.0764
LS change 0.0113 0.0050 0.0093 0.0063 0.0109 0.0071
income increase partner 15.1570 10.3008 14.1845 6.8101 22.4937 -1.2644
LS change 0.0152 0.0069 0.0175 0.0074 0.0229 0.0138
employment changes external 0.0977 0.1880 0.0687 0.1016 0.3131 0.0302
LS change 0.0000 -0.0007 -0.0006 -0.0004 -0.0007 -0.0006
single change external 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
LS change 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
health increase external 0.0472 0.0359 0.0474 0.0359 0.0530 0.0383
LS change 0.0000 -0.0035 -0.0021 -0.0021 -0.0018 -0.0019
income increase external 7.5785 5.0879 7.0422 3.3691 11.1796 -0.7040
LS change 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Reference income change 1.0000 1.0038 1.0035 1.0027 1.0041 1.0029
Reference health change 0.0000 0.0517 0.0310 0.0320 0.0277 0.0286
Reference employment change 0.0000 -0.0003 -0.0006 -0.0005 -0.0005 -0.0004
LS change 0.00000000 -0.00016609 -0.00010010 -0.00010925 -0.00008668 -0.00010177
All non-direct effects LS per patient 0.2563 0.1880 0.1563 0.1728 0.1894 0.1737
Overal LS effect per person 0.6193 0.4967 0.4531 0.4814 0.4980 0.4823
27. All non-direct effects LS per patient 0.2563 0.1881 0.1564 0.1729 0.1894 0.1738
Overal LS effect per person 0.6193 0.4967 0.4531 0.4814 0.4980 0.4823
Cost-Effectiveness Pounds per unit of LS 329.94
28.
29. Example 3
Adapted from:
Jody Ayres & John M. Malouff (2007)
Problem-solving training to help workers
increase positive affect, job satisfaction, and
life satisfaction, European Journal of Work and
Organizational Psychology, 16:3, 279-294,
DOI: 10.1080/13594320701391804
30.
31. Example 1. Basic background.
Airline Stewards in Australia have very
scripted and regulated jobs. When they fly,
where they fly, what they do, and how they do
it are all scripted. For some, this is stressful
and represents a lack of autonomy.
Researchers wanted to know if the stewards
would feel more in control and better about
their lives if they were given problem-solving
training. They convinced an Australian airline
to let them do it.
32. The training
111 volunteered for training and filled in
surveys before training and after training (4
weeks later). 56 got the training, 55 did not.
Training consisted of (in brief):
A 30 minute chat about life and life’s problems.
Homework whereby problems had to be written
down, and participants had to write down the key
elements of the problem, their strategy for solving
it, and why their solution worked or not.
Regular check-ups to see if the participants did this
and kept up their problem-solving diary.
33. Outcomes
In the 4 weeks, the ones with training improved 7.64 on a 1-50
score termed ‘positive affect’, relative to -0.2 for the control
group.
The intervention group improved 2.65 on a 5-35 score termed
‘life satisfaction’, relative to 0.82 for the control group.
The intervention group improved 1.57 on a 5-25 score termed
‘Job satisfaction’, relative to -0.11 for the control group.
The costs of this intervention were (derived from time-value
approximations) £148 per participant.
34. Standardised outcomes
The Work and Learning groups translate the measures in the
study to an improvement of 0.72 in standard 0-10 Life
Satisfaction and 0.56 in standard 0-10 Job Satisfaction.
The study thus finds that it costs £148 to improve Life
Satisfaction by 0.72 after 4 weeks, £202 per unit of
improvement at that time.
If one presumes the effect remains for a year (or halves every
8 months), this means the costs are £202 per unit of Life
Satisfaction per year.
If one assumes the effects last a month, one pays £2424 per
unit of Life Satisfaction per year. That is almost exactly on the
threshold of £2500 per year.
35. Notes on costs
Costs are for the intervention, not the study, which would add another £15
per participant.
The ‘participants’ now include those who got the training but were then lost
(they quit/were away). They were hence presumed to have had the same
benefit as those that filled in the forms after the training.
The costs are not based on what it cost in Australia, but what the same
time-investment would have been worth in the UK. Time of the stewards
(hourly earnings) and time of the ‘therapist’.
The hypothetical UK intervention is thus that of giving UK air travel
assistants the same training, conducted by an instructor without
institutional support (no one else involved, no set-up costs), paid for by the
stewards/airline via working fewer hours during the training.
If actual costs would be double and effects last a month, costs would be
£4848 per unit of LS per year. That would be above the threshold.
36. In this case…
Note: costs are in terms of production foregone and payments made.
This is not a public-expense lens (in which case the costs would be
taxes foregone and payments made).
Period MONTHS Period 1 Combined-undiscounted
Intended outcome PROBLEM-SOLVING 3.6 3.6
Direct cost per person 148 148
Other cost (+-) per person
External cost per person
Other outcome 1 0-10 JS 0.56 0.56
Other outcome 2 0-10 LS on externals
Direct LS effect on the treated per person 0-10 LS 0.72 0.72
Other LS effects per person
Overal LS effect per person 0-10 LS 0.72 0.72
Total internal cost per person
Total cost per person to society 148 148
Cost-effectiveness Pounds per unit of LS 205.56
37.
38. Hypothetical:
1. Start with the airline example of the main presentation. Then suppose you have
research showing the mood effect on an average individual has a 20% effect on the
mood of a partner, and that 80% of stewards have a partner.
Please fill in the relevant additional fields in the worksheet.
How does this change the cost-effectiveness of the study?
What key additional assumptions are you now making?
2. Suppose that relative job-conditions (including any training) are important amongst
stewards, and that 75% of all individual benefits from job-conditions come from their
job-conditions relative to others.
Please fill in the relevant additional fields in the worksheet.
How does this change the cost-effectiveness of the study?
What key additional assumptions are you now making?
3. (bonus: find on internet). How large is the likely effect of the improved job-
satisfaction on job-retention. And how much does it cost an average employer in that
kind of industry to fill a vacancy? How long do average jobs last? So how much is
likely saved in reduced vacancy costs due to the intervention?
Please fill in the relevant additional fields in the worksheet.
How does this change the cost-effectiveness of the study?
What key additional assumptions are you now making?
39. Suggested answers: on 1
Effectiveness is then 177 pounds per unit of LS
Key assumptions: stewards have representative relationships.
Period MONTHS Period 1 Combined-undiscounted
Intended outcome PROBLEM-SOLVING 3.6 3.6
Direct cost per person 148 148
Other cost (+-) per person
External cost per person 0 0
Other outcome 1 0-10 JS direct 0.56 0.56
Other outcome 2 0-10 LS on externals 0.1152 0.1152
Other outcome 3
Other outcome 4
Direct LS effect on the treated per person 0-10 LS 0.72 0.72
Other LS effects per person 0.1152 0.1152
Overal LS effect per person 0-10 LS 0.8352 0.8352
Total internal cost per person
Total cost per person to society 148 148
Cost-Effectiveness Pounds per unit of LS 177.20
40. On 2.
Effectiveness is then 709 pounds per unit of LS
Key assumptions: control group in the trial was not affected by the
changes to the small training group + reference effects are immediate
+ LS effects on partner now also reduce + JS effects also reduced.
Note that the losses are now on others who are not in the training.
Period MONTHS Period 1 Combined-undiscounted
Intended outcome PROBLEM-SOLVING 3.6 3.6
Direct cost per person 148 148
Other cost (+-) per person
External cost per person 0 0
Other outcome 1 0-10 JS direct 0.56 0.56
Other outcome 2 0-10 LS on externals 0.1152 0.1152
Other outcome 3 0-10 JS reference eff -0.42 -0.42
Other outcome 4 0-10 LS reference eff -0.6264 -0.6264
Direct LS effect on the treated per person 0-10 LS 0.72 0.72
Other LS effects per person -0.5112 -0.5112
Overal LS effect per person 0-10 LS 0.2088 0.2088
Total internal cost per person
Total cost per person to society 148 148
Cost-Effectiveness Pounds per unit of LS 708.81