Transcript Health Inequalities Research Programme

Data for planning equitable and
cost effective health services:
An approach from NZ
Burden of Disease Epidemiology, Equity & CostEffectiveness Programme (BODE3)
Directors: Tony Blakely, Nick Wilson, Diana Sarfati
Named Investigators: Hadorn, O’Dea, Tobias, McLeod, Costilla, Soeberg,
Atkinson, Simpson, Vos, Barendregt, Cobiac, Foster, Richardson, Sloane,
Kvizhinadze, Nghiem, Collinson
[email protected]
1
What I think I was meant to talk about
NZ Census-Mortality Study (NZCMS) and CancerTrends
• Nearly 25 years of mortality & cancer data linked to censuses
• Three examples of findings:
• Large undercounting of Māori and Pacific deaths and cancers in
1980s/90s, causing 20% to 35% underestimates of rates…
• … which when corrected for disclosed opening ethnic gaps in life
expectancy in the 1980s and 1990s [a time of structural reforms]
• Varying trends in cancer incidence over time, e.g. cervical cancer –
major public health success story!
85
80
75
70
65
60
55
50
45
1941
1951
1961
1971
Non-Māori Male
Māori Male
1981
1991
2001
Non-Māori Female
Māori Female
2
What is the next problem?
Policy making without synthesising evidence
• The 100 manila folder problem
• “Please sit on this committee, and advise us what to
do next.”
3
What are the opportunities?
Leverage existing data and methods
Unique ID linked health data
Census-mortality and census-cancer linked data
ACE methodology from Australia
Burden of disease studies – comparable disease
envelope and parameters
Increasing computer power
Data-banks of systematic reviews and meta-analyses
4
What is the idea?
Build infrastructure for rapid cost effectiveness analysis
Rather than respond to need for cost effectiveness
analyses one-by-one….
… first, build the data and modelling infrastructure that
can respond more rapidly and with greater
comparability between interventions to (just about)
anything you ask
Capitalise on New Zealand’s rich data by ethnicity and
socioeconomic position for equity analyses
Build capacity
5
Focus on economic decision models
Which is just one input into the decision making process
• Cost, effect (population change in health) and cost
effectiveness
• Equity
• Strength of the evidence base.
• Acceptability to stakeholders, especially public
• Feasibility of implementation
• Sustainability (Budget, workforce, political, other)
• Other consequences (side effects, spin-offs)
• Politics
• Social values
• Rule of rescue
6
Vision of BODE3
HRC-funded programme 2010-15; Ministry collaboration
“To build capacity and academic rigour in New
Zealand in the estimation of
disease burden, cost-effectiveness and equity
impacts of proposed interventions,
and undertake a range of such assessments.”
Burden of Disease Epidemiology, Equity &
Cost-Effectiveness Programme (BODE3)
uow.otago.ac.nz/BODE3-info.html
7
Yes, we will mainly use DALYs…
…. but in cost-effectiveness little different from QALYs
• There are two types of DALYs:
• For burden of disease studies where ‘external’ model
lifetable used [but no age weights a là 1990s GBD studies]
• For economic evaluations, where the population’s own
lifetable is used to determine background mortality rates
• Can talk in terms of ‘DALYs averted’, or ‘HALYs
gained’
• Thus the only conceptual difference is the use of
disability weights vis à vis utilities
8
Presentation
Structure of presentation to you today
• Objectives and methodologies for BODE3
• ABC-CBA
• NZACE-Prevention
• Building capacity and academic rigour
• Data inputs to infrastructure
• Interventions to assess
• Example of “link models”
9
2010 to 2015 objectives of BODE3
1. To estimate the impact and cost-effectiveness of cancer
control interventions
–
–
Markov time dependent macrosimulation models, and discrete event
simulation models
Aotearoa Burden of Cancer and Comparative Benefit Assessment
study; ABC-CBA
2. To estimate the impact and cost-effectiveness of preventive
interventions:
–
–
multistate lifetables
NZ-Assessing Cost-Effectiveness: Prevention; NZACE-Prevention.
3. To build capacity and academic rigour in
–
–
–
–
epidemiological and economic modelling
equity analyses
incorporation of uncertainty
skills and workforce
10
Integration of BODE3
ABC-CBA and NZACE-Prevention deliberately overlap
Injury
Diabetes
Burden of
Disease
CVD
Cancer
Palliative care
Treatment
ABC-CBA
Supportive care,
rehabilitation
Screening
Risk Factors
NZ-ACE Prevention
11
Objective 1: ABC-CBA
Capitalises on data strengths in New Zealand
INPUTS
Expert opinion
HealthTracker
(NHI linked data)
NZ data
DRG cost
estimates
Other cost data
Specify uncertainty distribution about
each input variable
Systematic
review of
literature
MODELLING
Interventionspecific
modelling of
change in core
model
parameters
(incidence,
survival, stage,
DW or utility)
Direct costing
of intervention
Societal
costing (if
appropriate)
Cancer model
Core disease
models
- Markov time
dependent
macrosimulation
- Discrete event
simulation
Attribution of
cancer Vote:Health
cost over Markov
states
OUTPUTS
Averted
disability
adjusted life
years (gained
HALYs)
Cost
effectiveness
Total change in
cost
12
Questions and Answers
Objective 1: ABC-CBA
Data used to build the baseline model
Current and future cancer incidence, by merging:
• Ministry of Health projections by sex by 5-year age group, with
• Linked census-cancer registration data (i.e. CancerTrends)
generated rate ratios of cancer
13
Questions and Answers
Breast cancer trends by ethnicity
Incidence c.f. mortality trends – census-linked data
Incidence from CancerTrends
Mortality from NZCMS
14
Index
Objective 1: ABC-CBA
Data used to build the baseline model
Current and future cancer incidence, by merging:
• Ministry of Health projections by sex by 5-year age group, with
• Linked census-cancer registration data (i.e. CancerTrends)
generated rate ratios of cancer
• Excess mortality rates (i.e. relative survival) from CancerTrends
–
–
•
Māori nearly always higher excess mortality (= lower relative survival)
A modest deprivation difference
Cost data from HealthTracker – Vote:Health costs assigned to
individuals (will also be used in NZACE-Prevention)
–
–
Vote:Health expenditure allocated across all individuals, by year, accounting
for up to 80% of Vote:Health budget
Use tabulations and regressions to generate ‘usual’ costs for a person with:
•
•
–
Given disease, or stage of cancer
Within year of death, within 6 months of diagnosis, etc…
These costs become the cost-offsets in economic decision models
15
Questions and Answers
Objective 1: ABC-CBA
Option 1: Time dependent Markov model:
Subpopulation
Maori Women age 45
in 2006
Cervical Cancer
Diagnosis &
Treatment
Remission
DW=0.25; 3 months
DW=0.20; variable time
Cure
After 5 years
Pre-terminal
DW=0.75; 5 months
Died of other causes
Terminal
DW=0.93; 1 month
Death
Questions and Answers
Objective 1: ABC-CBA
Actual modelling of interventions
•
•
Specify intervention
Parameterise in terms of:
–
–
–
–
–
•
Change in incidence rate
Change in survival
Change in stage distribution
Change in quality of life (be that DW or utility)
Change in direct costs (and possible ‘intervention-specific’ cost-offsets
downstream)
… often using ‘link models’ such as:
–
–
–
Care co-ordinators (or patient navigators) may hasten receipt of treatment,
which requires searching for literature on the impact of treatment ‘X’ weeks
earlier on survival chances, estimating ‘X’ for actual intervention, and
determining ‘change in survival’ (with uncertainty)
Event pathways for costing
Etc.
17
Questions and Answers
Objective 1: ABC-CBA
Early set of interventions to model
•
•
Selected with stakeholder advisory group; balance of relevance,
evidence, academic considerations
Initial set (biased to those with comparators, and equity interest):
–
–
–
–
Single versus multiple fraction radiotherapy for bone metastases
Docetaxel and paclitaxel for node positive breast cancer
Trastuzumab
Care co-ordinators (or patient navigators) for stage III colon cancer:
•
•
•
–
–
–
–
Diagnosis to surgery
Surgery chemotherapy
Adherence
Range of tobacco interventions (e.g. doubling calls to quitline)
Aspirin chemoprevention
CT screening for lung cancer
? Colorectal cancer screening programme
18
Questions and Answers
Objective 2: NZACE-Prevention
Assessing Cost-Effectiveness of Prevention
Overall aim:
To use an academically rigorous approach to “estimate the
disease burden impact and cost-effectiveness of preventive
interventions, for the population overall and by ethnicity and
socio-economic position”.
Uses multistate lifetables
Builds on ACE-Prevention Australia:
– Utilises existing and academically rigorous method
– … but will extend this work: context; interventions; methods.
Will use forthcoming New Zealand 2006 burden of disease study
parameters (from Ministry of Health)
19
Objective 2: NZACE-Prevention
Existing method; selecting of interventions
Focusing on six major risk factors (covering 38% of lost DALYs, all
relevant to inequalities) & have initially selected 91 interventions.
Risk factor
Examples of interventions to model
Tobacco use
Tobacco taxation increases, mass media campaigns, to expanding Quitline use and
providing new nicotine products for quitting.
High blood pressure
Reduction of salt in processed foods (voluntary and mandated options), to the
introduction of the polypill.
High cholesterol
Main initial focus, combining in absolute
risk approach, looking at fiscal policies
Promoting the use of food products with plant sterols to expanding the use of statins and
introduction of the (i.e.
polypill.taxes and subsidies)
Alcohol use
Alcohol taxation increases and alcohol advertising restrictions, to brief interventions
(by GPs).
Physical inactivity
Mass media-based campaigns and community programmes to encourage use of
pedometers, to a “green prescription” from a GP.
Overweight & obesity
Reduction of TV advertising (high fat/high sugar foods and drinks), to diet and physical
activity programmes.
20
Cost-effectiveness of alcohol interventions
ACE-Prevention (Australia), Cobiac et al
21
Questions and Answers
Obj. 3: Capacity and academic rigour
Methodological research
1. Equity analysis options – leverage off ‘heterogeneity’ of data.
– Separate modelling by social group
– Presenting DALYs-averted (HALYs-gained) by:
•
•
social group
targetted interventions
– We will trial measures of cost expressed per unit change in
absolute difference in per capita DALYs averted (HALYs gained)
– Equity-weighted benefit measures (e.g. equity weighted HALYs)
2. Uncertainty analyses:
– Parameter uncertainty routinely uses confidence intervals
– But systematic error often more important – we will develop
frameworks for incorporating systematic error
– Need for scenario analyses – not just mechanical PSA
3. Comparing DALYs & QALYs.
– Assessing the difference for an intervention that impacts on
disability/quality of life
22
BODE3: Current developments
Price elasticities as a complex example of ‘link models’
• Fiscal policies on food gaining momentum, e.g.:
• Danish fat tax
• Differential VAT by food type in Australia, and removing GST on
healthy food in New Zealand
• Requires ‘link models’:
• Tax/subsidy pass through rate:
• Wide range in literature; uncertainty
• Own-price elasticity:
• E.g. 1% increase in price of fruit leads to 0.6% decrease in
consumption (with uncertainty 0.3% to 1.0%)
• Cross-price elasticity:
• E.g. 1% increase in price of fruit leads to 0.1% increase in consumption
of (fatty, salty) potato crisps (with uncertainty …)
• Merging change in purchasing data with change in nutrient intake
• Specifying the change in nutrients with change in disease
23
BODE3: Current developments
Using expert knowledge
• All modelling requires ‘judgement’ or expert knowledge in
the specification of model structure
• Much modelling also requires expert knowledge in the
specification of input parameters (e.g. number of weeks a
care coordinator can hasten treatment by). There are
formal processes for this, e.g.:
• Expert panels
• Providing what information is known to panel members
• Asking them to estimate the most likely value and likely range (e.g.
interquartile) for true parameter
Leal et al. Eliciting Expert Opinion for Economic Models. Value in Health 2007;10(3):195-203.
O’Hagan A. Uncertain judgements. John Wiley and Sons, 2006
24
Data for planning equitable and
cost effective health services:
An approach from NZ
Burden of Disease Epidemiology, Equity & CostEffectiveness Programme (BODE3)
[email protected]
uow.otago.ac.nz/BODE3-info.html
uow.otago.ac.nz/cancertrends-info.html
uow.otago.ac.nz/nzcms-info.html
25