Comparative Effectiveness: UCSF East Africa Global Health -Kisumu 2014
1. UCSF Global Health Economics Colloquium
Cost-Effectiveness Workshop
Kisumu
20 January 2014
2. Purpose of CEA Workshop
• Basic understanding of CEA concepts &
methods
• Initial application to an issue of your
choosing
• Foundation for further development of ideas
and projects
2
3. Purpose of CEA
• “Opportunity cost” is a governing concept: resources
used for one purpose cannot be used for another.
• To foster efficient deployment of limited health
resources, we measure “opportunity costs”
• Assess the efficiency of available interventions to
achieve agreed health goals, e.g.,
– Less frequent vs. more frequent screening;
– Mobile vs. fixed facility service delivery;
– More vs. less intensive treatment.
• Examples from comparisons of interest to you
3
4. CEA Core Approach
• Incremental cost per standardized unit of health
gain
- E.g., per death averted or life-year gained
• For specified interventions; always compared with
other courses of action (standard of care, other
interventions)
• Is the inverse of (and equivalent to) health gain per
increment of spending
5. Key CEA outcome metric: ICER
• ICER – Incremental Cost-Effectiveness Ratio
• Δ costs / Δ health outcomes
– Δ means the difference between actions A and B
• ICER = [Cost A – Cost B] / [LifeYears A – LifeYears B]
• Incorrect: Cost A / LifeYears A. You need a comparator.
What are the incremental LYs or QALYs (or DALYs
averted)?
Option
Net cost
Δ cost
QALYs
Δ
QALYs
ICER
Drug A
$12,000
??
4.0
??
??
5
6. ICER Numerator
• Net costs = program costs adjusted for resulting
changes in medical costs
• Still, A vs B: net cost A – net cost B
• Medical costs can fall (averted disease) or rise
(identified or induced need for care)
6
7. ICER Denominator
• Difference between A and B in
o Natural health events (eg new infections or deaths
averted), or
o DALYs (disease burden, want to avert), or
o QALYs (health, want to gain)
7
8. “What is a „DALY’ anyway?”
• DALY = Disability-Adjusted Life-Years
• Summary measure of disease burden
• Sum of:
o Mortality (years lost due to premature death) +
o Morbidity (disability weight * duration in years).
• Opposite of QALY (measure of health)
• In global health DALYs used more than QALYs
QALY? DALY? Let’s call the whole thing off (someday)!
8
10. Disability weights: applications
For Descriptive Use (burden of disease):
Disability Weights must reflect the relative severity of
the consequences of different disease and disease
stages
Universal across time and over the globe
For Evaluative Use (cost-effectiveness)
Adjust time lived for level of disability from diseases
of interest other causes of disability
Measurement of non-fatal benefit of interventions
may involve modest changes in severity
More demands on accuracy of level of severity
11. Argument for using DALY DWs is that they are
derived through same process and are available
for a large number of diseases and health
states that are a consequence of disease
QALYs rely on utility weights: a difference between
two groups on a scale measuring the Quality of Life
is translated into a utility weight
QALY ≠ QALY ≠ QALY if utility weights are plucked
from disparate studies using different QoL
instruments and different methods of translating the
QoL scores into a utility value
12. Disability Weight: sources
1.
IHME Global Health Data Exchange (GHDx)
http://ghdx.healthmetricsandevaluation.org/record/global-burdendisease-study-2010-gbd-2010-disability-weights
14. A Basic CEA Results Table
Option
Net cost
Δ cost
DALYs
Δ
DALYs
ICER
($ per DALY
averted)
No
therapy
$10,000
n/a
4.0
n/a
n/a
Drug A
$12,000
?
3.5
?
?
Net cost = Cost of intervention
Adjusted for induced or averted health care costs
14
15. A Basic CEA Results Table
Option
Net cost
Δ cost
DALYs
Δ
DALYs
ICER
($ per DALY
averted)
No
therapy
$10,000
n/a
4.0
n/a
n/a
Drug A
$12,000
$2,000
3.5
0.5
$4,000
15
16. CEA Results Table – Negative ICER?!
Option
Net cost
Δ cost
DALYs
Δ
DALYs
(Δ $ / Δ DALY)
ICER
No therapy
$10,000
n/a
4.0
n/a
n/a
Drug A
$12,000
$2,000
3.5
0.5
$4,000
Drug B
$17,000
$5,000
?
3.75
- 0.25
?
- $20,000
?
16
17. CEA Results Table – “Dominance”
Option
Net cost
Δ cost
DALYs
Δ
DALYs
(Δ $ / Δ QALY)
ICER
No therapy
$10,000
n/a
4.0
n/a
n/a
Drug A
$12,000
$2,000
3.5
0.5
$4,000
Drug B
$17,000
$5,000
3.75
-0.25
Dominated
“Dominance” =
One option both cheaper and better than comparator
No trade-off = No brainer
Negative ICERs makes no sense.
17
18. CEA Table with Multiple Comparisons
Option
Net cost
Δ cost
DALYs
Δ
DALYs
(Δ $ / Δ DALY)
ICER
No therapy
$10,000
n/a
4.0
n/a
n/a
Drug A
$12,000
$2,000
3.5
0.5
$4,000
Drug B
$17,000
$5,000
3.75
-0.25
Dominated
Drug C
$18,000
$6,000
2.5
1.0
$6,000
Drug D
$23,000
$5,000
3.0
-0.5
Dominated
Array costs from lower to higher.
Compare each option to next higher (non-dominated) option.
Drug C vs. No therapy is inappropriate …
… feasible intermediate options must be evaluated.
No skipping allowed!
18
19. Introducing sensitivity analyses - Why do
them?
• All CEAs have substantial uncertainty.
• Sensitivity analyses deal with uncertainty systematically,
one input at a time and overall.
• Convince audience that results are robust (if you can).
• Show how results hinge on the value of certain inputs
• Show how key uncertainties, however disquieting
initially, actually do not affect findings in important ways.
Sensitivity analysis is mandatory in a CEA.
And interesting.
And fun!
19
20. One-way SA: Screening for gestational diabetes
Tornado diagram showing sensitivity of ICER to 16 key inputs. CCMH,
Chennai, India. Inputs varied 50% - 150% of base-case values
Marseille E et al. (2013). "The cost-effectiveness of gestational diabetes screening including prevention of type
2 diabetes: application of a new model in India and Israel." J Matern Fetal Neonatal Med.
Presentation title
Date
21. Multi-Way SA: Screening for gestational diabetes
20,000-trial Monte Carlo simulation, CCMH, Chennai, India. Distribution of ICER
values and 90% CI. Input values had beta distributions with minima and maxima at
50% and 150% of base case values
21
22. Analytic horizon – timing is everything
• What time period to portray?
– 30 days? 1 year? 5 years? A lifetime?
• Not standard … the rule is – capture important
differences between action options
– For treatment of a self-limited disease (i.e., trying to reduce
severity for a few weeks), perhaps 30 days.
– For an intervention with effects that decay by half each 6
months, perhaps 2-3 years
– For management of a chronic disease, perhaps lifetime.
• If in doubt, err toward longer time horizon … little extra
work.
22
24. Reference Case for CEA
• $ per QALY gained or DALY averted
• Societal (all payers), direct medical costs
• Discount future spending & health events, 3% per year
• Time horizon adequate to capture effects
• Report currency, price date, conversions
• Sensitivity analyses (evaluating uncertainty)
1. U.S. Preventive Services Task Force.
2. Consolidated Health Economic Evaluation Reporting
Standards (CHEERS) statement, BMJ, 2013.
24
26. CEA Example A – Research Questions
Cost-effectiveness of a mobile camp for adult male
circumcision in rural Zambia
• RQ1: What is the cost of delivering adult male
circumcision per 100 clients circumcised in this mobile
camp?
• RQ2: How many HIV infections and disability-adjusted
life years (DALYs) will be averted per 100 individuals
circumcised, in this population, over twenty years?
• RQ3: What is the incremental cost per DALY averted in
this population?
26
27. CEA Example A – RQ1
Cost-effectiveness of a mobile camp for adult male
circumcision in rural Zambia
• RQ1: What is the cost of delivering adult male
circumcision per 100 clients circumcised in this mobile
camp?
Brief methods: Review program financial and
service records for 12-month period, to quantify
resources used, associated costs, and clients
served.
27
28. CEA Example A – RQ2
Cost-effectiveness of a mobile camp for adult male
circumcision in rural Zambia
• RQ2: How many HIV infections and disability-adjusted
life years (DALYs) will be averted per 100 individuals
circumcised, in this population, over twenty years?
Brief methods: Build a decision analysis model
incorporating HIV epidemic projections with and
without circumcision.
28
29. CEA Example A – RQ3
Cost-effectiveness of a mobile camp for adult male
circumcision in rural Zambia
• RQ3: What is the incremental cost per DALY averted in
this population?
Brief methods: Calculate the incremental costeffectiveness ratio (ICER), with net costs (program
costs adjusted for changes in future HIV medical
care costs) in the numerator, and DALYs averted in
the denominator.
29
30. Introducing decision trees
• Graphically portrays the decision & its effects
• Three major components:
– The action options (the decision) under consideration.
– The probabilistic mix of consequences for each option.
– The value of health and cost outcomes for each consequence.
• Calculates the “expected value” for health and cost
outcomes for each option, as the weighted mean for the mix
of consequences.
30
31. A Basic Decision Tree
Voluntary adult male circumcision for HIV prevention in rural Kenya
HIV infection
0.4
No camp
No HIV infec.
0.6
Mobile circ.
camp?
# men
100
HIV infection
0.2
MC Camp
No HIV infec.
0.8
31
32. Adding Health
Outcomes
New HIV
infections
Per person
DALYs due
to new HIV
infections
7
HIV infection
0.4
40
280
No HIV infec.
0.6
0
0
HIV infection
0.2
20
140
No HIV infec.
0.8
0
0
No camp
Mobile circ.
camp?
# men
100
MC Camp
Infections averted DALYs averted
20
140
32
33. Adding
Intervention
Costs
New HIV
infections
Per person
HIV infection
0.4
40
DALYs due
to new HIV
infections
Cost of MC
Camp
7
$100
280
$0
No camp
No camp?
No costs!
No HIV infec.
0.6
0
$0
HIV infection
0.2
Mobile circ.
camp?
# men
100
0
20
140
$2,000
No HIV infec.
0.8
0
0
$8,000
MC Camp
Infections averted DALYs averted
20
140
Camp cost
$10,000
33
34. Adding
Medical Care
Costs
New HIV
infections
Per person
DALYs due
Cost of MC
to new HIV
Camp
infections
Cost of HIV
medical care
(if new HIV
infection)
7
$100
$6,000
$240,000
HIV infection
0.4
40
280
$0
No HIV infec.
0.6
0
0
$0
HIV infection
0.2
20
140
$2,000
No HIV infec.
0.8
0
0
$8,000
No camp
Mobile circ.
camp?
# men
100
$120,000
MC Camp
Infections averted DALYs averted
20
140
Camp cost
Med. Costs averted
$10,000
$120,000
34
35. And Finally
– Results!
New HIV
infections
Per person
DALYs due
Cost of MC
to new HIV
Camp
infections
Cost of HIV
medical care
(if new HIV
infection)
7
$100
$6,000
$240,000
Total cost
HIV infection
0.4
40
280
$0
No HIV infec.
0.6
0
0
$0
HIV infection
0.2
20
140
$2,000
No HIV infec.
0.8
0
0
$8,000
$240,000
No camp
Mobile circ.
camp?
# men
100
$0
$240,000
$120,000
$122,000
MC Camp
Infections averted DALYs averted
20
140
ICER ($/DALY averted)
$8,000
$130,000
Camp cost
Med. Costs averted
Net costs
$10,000
$120,000
($110,000)
Dominant
35
36. CEA Example B – Specific Aims
Cost-effectiveness of adherence counseling for HIV
anti-retroviral therapy (ART) in a primary care clinic in
rural India, as part of an RCT
• Specific Aim 1: Measure the cost in this primary care
setting of adherence counseling per client receiving the
service and per patient-year of ART.
• Specific Aim 2: Estimate the impact of adherence
counseling on HIV disease progression and disabilityadjusted life years (DALYs) over three and ten years.
• Specific Aim 3: Calculate the cost per added individual
with viral suppression and the cost per DALY averted.
36
37. CEA Example B – Specific Aim 1
Cost-effectiveness of adherence counseling for HIV
anti-retroviral therapy (ART) in a primary care clinic in
rural India, as part of an RCT
• Specific Aim 1: Measure the cost in this primary care
setting of adherence counseling per client receiving the
service and per patient-year of ART.
Brief methods: Review program financial and service
records for a 12 month period, in order to quantify
adherence counseling costs (resources used and associated
costs) and client-years of ART. Use time and motion
methods to separate staff effort dedicated to adherence
counseling from other activities.
37
38. CEA Example B – Specific Aim 2
Cost-effectiveness of adherence counseling for HIV
anti-retroviral therapy (ART) in a primary care clinic in
rural India, as part of an RCT
• Specific Aim 2: Estimate the impact of adherence
counseling on HIV disease progression and disabilityadjusted life years (DALYs) over three and ten years.
Brief methods: Build a decision analysis to portray HIV
disease progression on ART as a function of viral
suppression, and associated DALYs due to premature
mortality and morbidity. Set model values for three years
from RCT results, and project to ten years using disease
state modeling.
38
39. CEA Example B – Specific Aim 3
Cost-effectiveness of adherence counseling for HIV
anti-retroviral therapy (ART) in a primary care clinic in
rural India, as part of an RCT
• Specific Aim 3: Calculate the cost per added individual
with viral suppression and the cost per DALY averted.
Brief methods: Compare program costs to RCT measures
of rate of viral suppression. Calculate the ICER ($ per DALY
averted) with net costs (adherence counseling costs
adjusted for changes in ART costs) in the numerator, and
DALYs averted in the denominator.
39
40. Cost-effectiveness of adherence counseling for ART, primary care clinic in rural India, in RCT
Simplified tree
Health outcomes
Deaths on ART
(3 years)
Per person
Death
0.2
DALYs
(death;
alive)
10
4
Health outcomes
Cost of
Cost of HIV
Adherence medical care Total cost
Counseling (death; alive)
$30
$1,200
$7,000
0.2
2
$240
$240
0.8
0.8
3.2
5.2
$5,600
$5,600
$5,840
Death
0.15
0.15
1.5
$5
$180
$185
Alive
0.85
0.85
3.4
4.9
$26
$5,950
$5,976
$6,160
DALYs averted
0.3
Cost
$320
No
Alive
Adherence
counseling?
# patients
1
Yes
Comparing
ICER ($/DALY averted
$1,067
40
41. What cost-effectiveness research
question are you interested in?
• What is:
– the intervention?
– the comparator?
• What outcome measures are appropriate?
• How will you evaluate intervention benefits?
• How will you measure program costs?
• Will you adjust for changes in direct medical costs
resulting from the intervention?
• Can you sketch a tree that portrays the consequences of
the intervention and its comparator?
41
Editor's Notes
This dataset provides the disability weights with uncertainty intervals for 220 unique health states used to estimate non-fatal health outcomes in the GBD 2010 Study. The data were published in The Lancet in December 2012 in the paper “Common values in assessing health outcomes from disease and injury: disability weights measurement study for the Global Burden of Disease Study 2010.”