View
215
Download
1
Category
Tags:
Preview:
Citation preview
Modelling Drug Policies
Gabriele BammerNational Centre for Epidemiology and Population Health,
The Australian National UniversityHauser Center for Nonprofit Organizations,
Harvard University
Presentation overview
Where does modelling fit? Modelling as a tool for
descriptionexplanationpredictionimputationintegrationevaluation
Problem framing and boundary setting Ignorance and uncertainty Capacity Integration and Implementation Sciences
Where does modelling fit?...1
Move away from narrowly defined and delimited problems to complex problems requiring multiple inputs – less well developed
Aid to thinking – conceptual ± quantification
“A schematic description of a system, theory, or phenomenon that accounts for its known or inferred properties and may be used for further study of its characteristics: a model of generative grammar; a model of an atom; an economic model.” (http://www.thefreedictionary.com/modelling).
Where does modelling fit?...2
How can modelling help drugs research? Modelling as a tool for
description
explanation
prediction
imputation
integration
evaluation
(usually more than one; illustrate multiple ways model can be used)
Where does modelling fit?...3
Broad overview - no specifics, but flavour, plus strengths of some particular types of modelling
Not typology Not comprehensive Bigger is not always better Never perfect representation of reality
- framing and boundaries- ignorance and uncertainty (capacity)
My approach- modelling as a view of a system- modelling sceptic
Presentation overview
Where does modelling fit? Modelling as a tool for
descriptionexplanationpredictionimputationintegrationevaluation
Problem framing and boundary setting Ignorance and uncertainty Capacity Integration and Implementation Sciences
Modelling for description...1
Ability to show connections between elements in problem of interest
Appreciate system in which model is embedded
Example
Model of tobacco control developed by a US National Cancer Institute project “Initiative for the Study and Implementation of Systems”
Modelling for description...2
Funding for tobaccocontrol programs
Gov. incomefrom tabacco
taxes
Tobacco controlprograms
Smokers
Perceived importance tofocus on other health
programs
Public awarenessof tobaccohealth risk
Pressure on tobaccocompanies to reducemarketing activities
Tobacco marketingactivities
Taxrevenues
fromsmokers
+
+
+
-
+
People quittingsmoking
-
Fraction of peoplesmoking
Smoking as asocial norm
People startingsmoking
Tobaccorevenues
+
+ +
+
++
+
Health care costs
+
Health insurerscoverage of tobacco
quitting costs
+
+
Researchersawarness of
tobacco healthrisk
Funding fortobacco health
research+
+
Govt awarenessof tobaccohealth risk
+
-
Pro-tobaccocontituencies
Anti-tobaccoconstituencies
++ +
+
Tobacco productsavailability
+
Tobaccogrowers
+
+
+
++
+
Govt willingness tolegislate tobacco
control
- + +
Tobaccotaxes
Govt funding oftobacco control
--
Trend in tobaccocompany revenues
+
-
Anti-smokinglegislation
-
Modelling for description…3
Advantage of system dynamics models: considers what those linkages mean in terms of
increasing or decreasing the elements (in system dynamics terms ‘stocks’)
Funding for tobaccocontrol programs
Gov. incomefrom tabacco
taxes
Tobacco controlprograms
Smokers
Perceived importance tofocus on other health
programs
Public awarenessof tobaccohealth risk
Pressure on tobaccocompanies to reducemarketing activities
Tobacco marketingactivities
Taxrevenues
fromsmokers
+
+
+
-
+
People quittingsmoking
-
Fraction of peoplesmoking
Smoking as asocial norm
People startingsmoking
Tobaccorevenues
+
+ +
+
++
+
Health care costs
+
Health insurerscoverage of tobacco
quitting costs
+
+
Researchersawarness of
tobacco healthrisk
Funding fortobacco health
research+
+
Govt awarenessof tobaccohealth risk
+
-
Pro-tobaccocontituencies
Anti-tobaccoconstituencies
++ +
+
Tobacco productsavailability
+
Tobaccogrowers
+
+
+
++
+
Govt willingness tolegislate tobacco
control
- + +
Tobaccotaxes
Govt funding oftobacco control
--
Trend in tobaccocompany revenues
+
-
Anti-smokinglegislation
-
Funding for tobaccocontrol programs
Gov. incomefrom tabacco
taxes
Tobacco controlprograms
Smokers
Perceived importance tofocus on other health
programs
Public awarenessof tobaccohealth risk
Pressure on tobaccocompanies to reducemarketing activities
Tobacco marketingactivities
Taxrevenues
fromsmokers
+
+
+
-
+
People quittingsmoking
-
Fraction of peoplesmoking
Smoking as asocial norm
People startingsmoking
Tobaccorevenues
+
+ +
+
++
+
Health care costs
+
Health insurerscoverage of tobacco
quitting costs
+
+
Researchersawarness of
tobacco healthrisk
Funding fortobacco health
research+
+
Govt awarenessof tobaccohealth risk
+
-
Pro-tobaccocontituencies
Anti-tobaccoconstituencies
++ +
+
Tobacco productsavailability
+
Tobaccogrowers
+
+
+
++
+
Govt willingness tolegislate tobacco
control
- + +
Tobaccotaxes
Govt funding oftobacco control
--
Trend in tobaccocompany revenues
+
-
Anti-smokinglegislation
-
Presentation overview
Where does modelling fit? Modelling as a tool for
descriptionexplanationpredictionimputationevaluation
Problem framing and boundary setting Ignorance and uncertainty Capacity Integration and Implementation Sciences
Modelling for explanation...1
Move beyond description to understanding what is going on
Examples: Revisit system dynamics model West’s synthetic model of addiction
Funding for tobaccocontrol programs
Gov. incomefrom tabacco
taxes
Tobacco controlprograms
Smokers
Perceived importance tofocus on other health
programs
Public awarenessof tobaccohealth risk
Pressure on tobaccocompanies to reducemarketing activities
Tobacco marketingactivities
Taxrevenues
fromsmokers
+
+
+
-
+
People quittingsmoking
-
Fraction of peoplesmoking
Smoking as asocial norm
People startingsmoking
Tobaccorevenues
+
+ +
+
++
+
Health care costs
+
Health insurerscoverage of tobacco
quitting costs
+
+
Researchersawarness of
tobacco healthrisk
Funding fortobacco health
research+
+
Govt awarenessof tobaccohealth risk
+
-
Pro-tobaccocontituencies
Anti-tobaccoconstituencies
++ +
+
Tobacco productsavailability
+
Tobaccogrowers
+
+
+
++
+
Govt willingness tolegislate tobacco
control
- + +
Tobaccotaxes
Govt funding oftobacco control
--
Trend in tobaccocompany revenues
+
-
Anti-smokinglegislation
-
Modelling for explanation...3
Move beyond description to understanding what is going on
Examples: Revisit system dynamics model West’s synthetic model of addiction
Responses (starting, modifying or stopping an action)
Impulses/inhibitions (urges)
Motives (wants/needs)
Evaluations (beliefs) PlansEmotional states (e.g. happiness
distress)
Drives (e.g. hunger)
Stimuli/information
Arousal
West’s synthetic model of addiction (2006)
Presentation overview
Where does modelling fit? Modelling as a tool for
descriptionexplanationpredictionimputationintegrationevaluation
Problem framing and boundary setting Ignorance and uncertainty Capacity Integration and Implementation Sciences
Modelling for prediction...1
Help appreciate likely consequences of particular actions – huge benefit of models
But BEWARE!!!
Paradox: surprising results – value and danger
Examples: Revisit system dynamics modelDPMP- Australian heroin market
Also more complex economic micro-simulation models
Modelling for prediction...2
Youthnonsmokers
Adultnonsmokers
Youthsmokers
Adultsmokers
Youthformer
smokersAdultformer
smokers
Aging 1 Adult neversmokers dying
Youth smokingstarts
Adult smokingstarts
Youth quitsYouth formersmokers smoking Adult quits
Time to age 1Nonsmoking adult
lifespan
Norm yth uptake Norm adult uptake
Norm youth quitrate
Norm adult quitrate
Youth returning tosmoking
Adult returning tosmoking
Adult formersmokers smoking
Youth smokersaging
Former youthsmokers aging
Former adultsmokers dying
Smoking adultlifespan
Adult smokersdying
Norm former adultsmoker lifespan
Time to age 2
Time to age 3
Effect of public opinionon adult quit rate
Effect of public opinionon adult uptake
Effect of public opinionon youth uptake
Effect of public opinionon youth quit rate
<Total adults>
New youth
Norm birth rate
Note. Years are the unit of age. yth = youth.
Model of Aging Chains of Smokers (Birth to Death)
Modelling for prediction...3
Adults smoking40 M
35 M
30 M
25 M
20 M
1980 1990 2000 2010 2020Time (Year)
Adults smoking : cut adult init in half peopleAdults smoking : cut teen init in half peopleAdults smoking : cut child init in half peopleAdults smoking : smoking chain base people
Modelling for prediction...4
Help appreciate likely consequences of particular actions – huge benefit of models
But BEWARE!!!
Paradox: surprising results – value and danger
Examples: Revisit system dynamics modelDPMP- Australian heroin market (Moore et al, 2005)
Also more complex economic micro-simulation models
Modelling for prediction...4
Economic return from
dealing =
Revenue from selling drugs
-
Cost of obtain-ing the drugs sold
-
Convent-ional
business costs
-Non-costs
Presentation overview
Where does modelling fit? Modelling as a tool for
descriptionexplanationpredictionimputationintegrationevaluation
Problem framing and boundary setting Ignorance and uncertainty Capacity Integration and Implementation Sciences
Modelling for imputation
Problems in drugs area of lack of data
Imputation bridges data gaps – use in demography for mortality estimates in developing countries
Examples: DPMP, estimates of heroin users using truncated Poisson
and Poisson regression modelling for the former and the application of a Jolly-Seber type model (Dietze et al, 2005)
Good example of not trusting results
Presentation overview
Where does modelling fit? Modelling as a tool for
descriptionexplanationpredictionimputationintegrationevaluation
Problem framing and boundary setting Ignorance and uncertainty Capacity Integration and Implementation Sciences
Modelling for integration...1
Model is an additional benefit – main benefit is that it provides a focus for bringing together different disciplinary and stakeholder perspectives
Examples:
Return to tobacco control
DPMP: complex adaptive systems
Modelling for integration...3
Model is an additional benefit – main benefit is that it provides a focus for bringing together different disciplinary and stakeholder perspectives
Examples:Return to tobacco controlDPMP: complex adaptive systems (Perez et al., 2005)
Drug use in Melbourne:Epidemiology, treatment, crime, economics
Includes dynamics
Presentation overview
Where does modelling fit? Modelling as a tool for
descriptionexplanationpredictionimputationintegrationevaluation
Problem framing and boundary setting Ignorance and uncertainty Capacity Integration and Implementation Sciences
Modelling for evaluation...1
Theory-based Evaluation or Program Logic or Program Theory
Makes assumptions explicit Links program strategy to evaluation Systematic link between inputs and activities and
outcomes
Example:Evaluation of WA community crime reduction program
(Cummings 2005)
Modelling overall
No model can provide a perfect representation of reality
Every model has significant limitations
Move on now to three key limitations
Presentation overview
Where does modelling fit? Modelling as a tool for
descriptionexplanationpredictionimputationintegrationevaluation
Problem framing and boundary setting Ignorance and uncertainty Capacity Integration and Implementation Sciences
Problem framing and boundary setting…1
Importance of framing political ‘spin’ ‘junkie’, ‘drug user’, ‘treatment client’ and ‘pretty
criminal’
Examples:Drug use as a problem or as a symptomDrugs and Public Policy Project: agent-host-vector
approachPollack: symbolism of needle-syringe exchange
Problem framing and boundary setting…2
Boundary setting : Attention to what is included, excluded and marginalised- link with framing- link with values
Example:Model of tobacco control:
Exclude industry perspectiveMarginalise smokers (assume advocates for tobacco industry, therefore not consulted)
Methods include:Scoping methodsUlrich’s Critical system heuristics
Scoping…1
Scoping is the preparatory stage of a project where we systematically think about what we can best do with the time, money and person-power we have at our disposal.
It involves considering:what we want to achieve and who will be affectedwho needs to be on sidewhat needs to be done to get therewhat are the likely blocks and how they can be overcome.
Scoping…2
Helps us: Broaden our view of the problem beyond what we
know and understand, recognising and respecting different points of view
Set boundaries Decide if we want to challenge the way the problem
is generally viewed, by paying more attention to something society sees as marginal or excluded
Work out who we need to have on-side to give the project legitimacy.
Scoping…3
Eight questions useful for scoping:1. What do we know about the problem?2. What can different interest groups and academic
disciplines contribute to addressing this problem?3. What areas are contentious?4. What are the big picture issues? In other words,
what are the political, social and cultural aspects of the problem?
These tackle the dimensions of the problem.
Scoping…4
5. Why is this problem on the agenda now?
6. What support and resources are likely to be available for tackling the problem?
7. What parts of the problem are already well covered and where are the areas of greatest need?
8. Where can the most strategic interventions be made?
These tackle the priorities.
Presentation overview
Where does modelling fit? Modelling as a tool for
descriptionexplanationpredictionimputationintegrationevaluation
Problem framing and boundary setting Ignorance and uncertainty Capacity Integration and Implementation Sciences
Ignorance and uncertainty…1
Understanding is comparatively unsophisticated
No effective handles
Different disciplines and practice areas
- different approaches
- different importance
Nobody’s mandate to pull different approaches together
Ignorance and uncertainty…2
Primary level
Meta-level
Known Unknown
Known Known knowns
Unknown knowns (tacit knowledge)
Unknown Known unknowns (conscious ignorance)
Unknown unknowns (meta-ignorance)
Ignorance and uncertainty…3
STATISTICS - probability theory
Intelligence – gaps or overloadHistory – moral dimension
Music – essential for creativity
Art – certainty and uncertainty are a continuum, not opposites
Futures – unknown unknownsComplexity - irreducible
Religion – desirable vs fundamentalism
Presentation overview
Where does modelling fit? Modelling as a tool for
descriptionexplanationpredictionimputationintegrationevaluation
Problem framing and boundary setting Ignorance and uncertainty Capacity Integration and Implementation Sciences
Capacity…1
Watch out – modelling tends to take on a life of its own
Need to worry about limited research capacity and how to use it wisely
especially on a global level
Institute of Scientific Information publications listings
Capacity…2
Country Affiliations of Authors of Papers Listed in Web of Science (Jan. 1998 - Mar. 2003)
0
5,000
10,000
15,000
20,000
25,000
Low-Income Lower-Middle-Income
Upper-Middle-Income
Upper-Income
Economies
Au
tho
rs (
Me
dia
n)
High: 104,865 (India)
Median: 236
High: 153,753 (Russian
Federation)Median: 737
High: 60,207 (Poland)Median: 1,375
High: 2,060,522 (USA)Median: 19,169
Presentation overview
Where does modelling fit? Modelling as a tool for
descriptionexplanationpredictionimputationintegrationevaluation
Problem framing and boundary setting Ignorance and uncertainty Capacity Integration and Implementation Sciences
Integration and Implementation Sciences…1
Modelling as an integrative tool Modelling for decision support – all of the
above Problem framing and boundary setting Ignorance and uncertainty (Capacity)
Integration and Implementation Sciences...2
Concepts and methods to improve: the generation of knowledge spanning disciplines
and practice,
Integration and Implementation Sciences...3
Concepts and methods to improve: Knowledge generation,
the application of that knowledge in decision making
in policy, business, professional practice and community activism,
Integration and Implementation Sciences...4
Concepts and methods to improve: knowledge generation, decision making, and
the implementation of those decisions to bring about effective change and social improvement.
Integration and Implementation Sciences?...6
No existing toolkit for methods to improve integration
ModelsDialogue
Common metric
Product
‘Vision’
Integration and Implementation Sciences...7
Research to improve decision support - all forms of modelling are useful
Modelling as a tool for
description
explanation
prediction
imputation
integration
evaluation
Key concepts
Better ways of dealing with:
Ignorance and uncertainty Problem framing and boundary setting Values definition Collaboration
Why do we need it?
Recognised urgency to tackle complex problems
Learning by doing
No cross-fertilisation (fragmentation)
Reinventing of the wheel
Lack of quality control
No formal methodology or training (statistics analogy)
Need clear and systematic way for building integration into research projects
Cross-cutting specialisation
Application in a specific sector
Methodological
development with respect to a single sector
ENVIRONMENTHEALTH
INNOVATION & BUSINESS
RISK & SECURITY
INTERNATIONAL DEVELOPMENT
EDUCATION
POLICY & GOVERNMENT
SOCIETY, HUMAN BEHAVIOUR & CULTURE
Theory and
methods
Conclusions
Modelling has a lot to offer drugs research: description, explanation, prediction, imputation, integration, evaluation
Different modelling methods have different strengths
Every model has limitations
Case for a new specialisation – Integration and Implementation Sciences
What do you think?
Would this help you in your work?
Acknowledgements
Funding: Fulbright, Colonial Foundation Trust, NHMRC, LWA, Hauser Center, NCI (ISIS), GECAFS
Tobacco control: George RichardsonImputation: Terry HullEvaluation: David McDonald
DPMP – Alison Ritter, Gerald Midgley, Wendy Gregory, Pascal Perez, Anne Dray
Uncertainty – Michael Smithson
Recommended