Syndemics
Prevention Network
Overview of System Dynamics Simulation Modeling
Systems Thinking and Modeling WorkshopOffice of Disease Prevention and Health Promotion
Bethesda, MDMay 8, 2006
Bobby MilsteinSyndemics Prevention Network
Centers for Disease Control and PreventionAtlanta, Georgia
Syndemics
Prevention Network
Research Imperatives for Protecting Health
Gerberding JL. Protecting health: the new research imperative. Journal of the American Medical Association 2005;294(11):1403-1406.
Typical Current StateStatic view of problems that are studied in isolation
Proposed Future StateDynamic systems and syndemic approaches
"Currently, application of complex systems theories or syndemic science
to health protection challenges is in its infancy.“
-- Julie Gerberding
Syndemics
Prevention Network
1999 2000 2001 2002 2003 2004 2005
System Change Initiatives Encounter Limitations of Logic Models and
Conventional Planning/Evaluation Methods
Diabetes Action Labs*
ODPHP Modelers Meeting
Upstream-Downstream Investments
Obesity Overthe Lifecourse*
Fetal & Infant Health Goal-Setting
Milestones in the Recent Use of System Dynamics Modeling at CDC
AJPH Systems
Issue
2006
CDC Evaluation Framework
Recommends Logic Models
SD Emerges as a Promising Methodology
Neighborhood Assistance
Game
HypertensionPrevention &
Control *
Syndemics Modeling
Science Seminars and Professional Development Efforts
* Dedicated multi-year budget
Syndemics
Prevention Network
System Dynamics Was Designed to Address Problems Marked By Dynamic Complexity
Good at Capturing
• Differences between short- and long-term consequences of an action
• Time delays (e.g., transitions, detection, response)
• Accumulations (e.g., prevalence, capacity)
• Behavioral feedback (e.g., actions trigger reactions)
• Nonlinear causal relationships (e.g., effect of X on Y is not constant)
• Differences or inconsistencies in goals/values among stakeholders
Sterman JD. Business dynamics: systems thinking and modeling for a complex world. Boston, MA: Irwin McGraw-Hill, 2000.
Homer JB, Hirsch GB. System dynamics modeling for public health: background and opportunities. American Journal of Public Health 2006;96(3):452-458.
Origins
• Jay Forrester, MIT (from late 1950s)
• Public policy applications starting late 1960s
Syndemics
Prevention Network
Understanding Dynamic ComplexityFrom a Very Particular Distance
“{System dynamics studies problems} from ‘a very particular distance', not so close as to be concerned with the action of a single individual, but not so far away
as to be ignorant of the internal pressures in the system.”
-- George Richardson
Forrester JW. Counterintuitive behavior of social systems. Technology Review 1971;73(3):53-68.
Meadows DH. Leverage points: places to intervene in a system. Sustainability Institute, 1999. Available at <http://www.sustainabilityinstitute.org/pubs/Leverage_Points.pdf>.
Richardson GP. Feedback thought in social science and systems theory. Philadelphia, PA: University of Pennsylvania Press, 1991.
Sterman JD. Business dynamics: systems thinking and modeling for a complex world. Boston, MA: Irwin McGraw-Hill, 2000.
Syndemics
Prevention Network
Time Series Models
Describe trends
Multivariate Stat Models
Identify historical trend drivers and correlates
Patterns
Structure
Events
Increasing:
• Depth of causal theory
• Degrees of uncertainty
• Robustness for longer-term projection
• Value for developing policy insights
Increasing:
• Depth of causal theory
• Degrees of uncertainty
• Robustness for longer-term projection
• Value for developing policy insights
Dynamic Simulation Models
Anticipate new trends, learn about policy consequences,
and set justifiable goals
Tools for Policy Analysis
Syndemics
Prevention Network
Wickelgren I. How the brain 'sees' borders. Science 1992;256(5063):1520-1521.
How Many Triangles Do You See?
Syndemics
Prevention Network
Ulrich W. Reflective practice in the civil society: the contribution of critically systemic thinking. Reflective Practice 2000;1(2):247-268. http://www.geocities.com/csh_home/downloads/ulrich_2000a.pdf
Boundary Critique
Syndemics
Prevention Network
Ulrich W. Reflective practice in the civil society: the contribution of critically systemic thinking. Reflective Practice 2000;1(2):247-268. http://www.geocities.com/csh_home/downloads/ulrich_2000a.pdf
Boundary Critique
Syndemics
Prevention Network
Milstein B, Homer J. The dynamics of upstream and downstream: why is so hard for the health system to work upstream, and what can be done about it? CDC Futures Health Systems Workgroup; Atlanta, GA; 2003.
TertiaryPrevention
SecondaryPrevention
PrimaryPrevention
TargetedProtection
Society's HealthResponse
Demand forresponse
PublicWork
SaferHealthierPeople Becoming
vulnerable
Becoming saferand healthier
VulnerablePeople Becoming
afflicted
Afflictedwithout
Complications Developingcomplications
Afflicted withComplications
Dying fromcomplications
Health System Dynamics
Adverse LivingConditions
GeneralProtection
Milstein B, Homer J. The dynamics of upstream and downstream: why is so hard for the health system to work upstream, and what can be done about it? CDC Futures Health Systems Work Group; Atlanta, GA; December 3, 2003.
Gerberding JL. CDC's futures initiative. Atlanta, GA: Public Health Training Network; April 12, 2004.
Homer JB, Hirsch GB. System dynamics modeling for public health: background and opportunities. American Journal of Public Health 2006;96(3):452-458.
Syndemics
Prevention Network
Understanding Health as Public Work
SaferHealthierPeople
VulnerablePeople
Afflictedwithout
Complications
Afflicted withComplicationsBecoming
vulnerable
Becoming saferand healthier
Becomingafflicted
Developingcomplications
Dying fromcomplications
Adverse LivingConditions
Society's HealthResponse
Demand forresponse
GeneralProtection
TargetedProtection
PrimaryPrevention
SecondaryPrevention
TertiaryPrevention
-
Public Work-
Vulnerable andAfflicted People
Fraction of Adversity,Vulnerability and AfflictionBorne by Disadvantaged
Sub-Groups (Inequity)
PublicStrength
-
Citizen Involvementin Public Life
Social Division
Syndemics
Prevention Network
Testing Dynamic Hypotheses
-- How can we learn about the consequences of actions in a system of this kind?-- Could the behavior of this system be analyzed using conventional epidemoiological methods (e.g., logistic or multi-level regression)?
SaferHealthierPeople
VulnerablePeople
Afflictedwithout
Complications
Afflicted withComplicationsBecoming
vulnerable
Becoming saferand healthier
Becomingafflicted
Developingcomplications
Dying fromcomplications
Adverse LivingConditions
Society's HealthResponse
Demand forresponse
GeneralProtection
TargetedProtection
PrimaryPrevention
SecondaryPrevention
TertiaryPrevention
-
Public Work-
Vulnerable andAfflicted People
Fraction of Adversity,Vulnerability and AfflictionBorne by Disadvantaged
Sub-Groups (Inequity)
PublicStrength
-
Citizen Involvementin Public Life
Social Division
Syndemics
Prevention Network
Learning In and About Dynamic Systems
Benefits of Simulation/Game-based Learning
• Formal means of evaluating options
• Experimental control of conditions
• Compressed time
• Complete, undistorted results
• Actions can be stopped or reversed
• Visceral engagement and learning
• Tests for extreme conditions
• Early warning of unintended effects
• Opportunity to assemble stronger support
Dynamic Complexity Hinders…
• Generation of evidence (by eroding the conditions for experimentation)
• Learning from evidence (by demanding new heuristics for interpretation)
• Acting upon evidence (by including the behaviors of other powerful actors)
Sterman JD. Learning from evidence in a complex world. American Journal of Public Health (in press).
Sterman JD. Business dynamics: systems thinking and modeling for a complex world. Boston, MA: Irwin McGraw-Hill, 2000.
“In [dynamically complex] circumstances simulation becomes the only reliable way to test a hypothesis and evaluate the likely effects of policies."
-- John Sterman
Syndemics
Prevention Network
System Dynamics Modeling SupportsNavigational Policy Dialogues
Prevalence of Diagnosed Diabetes, US
0
10
20
30
40
1980 1990 2000 2010 2020 2030 2040 2050
Mill
ion
pe
op
le
HistoricalData
Markov Model Constants• Incidence rates (%/yr)• Death rates (%/yr)• Diagnosed fractions(Based on year 2000 data, per demographic segment)
Honeycutt A, Boyle J, Broglio K, Thompson T, Hoerger T, Geiss L, Narayan K. A dynamic markov model for forecasting diabetes prevalence in the United States through 2050. Health Care Management Science 2003;6:155-164.
Jones AP, Homer JB, Murphy DL, Essien JDK, Milstein B, Seville DA. Understanding diabetes population dynamics through simulation modeling and experimentation. American Journal of Public Health 2006;96(3):488-494.
Why?
Where?
How?
Who?
What?
Markov Forecasting Model
Simulation Experiments
in Action Labs
Syndemics
Prevention Network
Simulations for Learning in Dynamic Systems
Morecroft JDW, Sterman J. Modeling for learning organizations. Portland, OR: Productivity Press, 2000.
Sterman JD. Business dynamics: systems thinking and modeling for a complex world. Boston, MA: Irwin McGraw-Hill, 2000.
Sterman JD. Learning from evidence in a complex world. American Journal of Public Health 2006;96(3):505-514.
Sterman JD. All models are wrong: reflections on becoming a systems scientist. System Dynamics Review 2002;18(4):501-531.
Multi-stakeholder Dialogue
Dynamic Hypothesis (Causal Structure) Plausible Futures (Policy Experiments)Deaths per Population
0.0035
0.003
0.0025
0.002
0.0015
1980 1990 2000 2010 2020 2030 2040 2050
Time (Year)
Blue: Base run; Red: Clinical mgmt up from 66% to 90%;Green: Caloric intake down 4% (99 Kcal/day);Black: Clin mgmt up to 80% & Intake down 2.5% (62 Kcal/day)
Base
Downstream
Upstream
Mixed
“All models are wrong. Some are useful.”
Syndemics
Prevention Network
“Simulation is a third way of doing science.
Like deduction, it starts with a set of explicit
assumptions. But unlike deduction, it does not
prove theorems. Instead, a simulation generates
data that can be analyzed inductively. Unlike
typical induction, however, the simulated data
comes from a rigorously specified set of rules
rather than direct measurement of the real world.
While induction can be used to find patterns in
data, and deduction can be used to find
consequences of assumptions, simulation
modeling can be used as an aid to intuition.”
-- Robert Axelrod
Axelrod R. Advancing the art of simulation in the social sciences. In: Conte R, Hegselmann R, Terna P, editors. Simulating Social Phenomena. New York, NY: Springer; 1997. p. 21-40. <http://www.pscs.umich.edu/pub/papers/AdvancingArtofSim.pdf>.
Sterman JD. Business Dynamics: Systems Thinking and Modeling for a Complex World. Boston, MA: Irwin McGraw-Hill, 2000.
Simulation ExperimentsOpen a Third Branch of Science
“The complexity of our mental models vastly exceeds our ability to understand their implications without simulation."
-- John Sterman
How?
Where?
0
10
20
30
40
50
1960-62 1971-74 1976-80 1988-94 1999-2002
Prevalence of Obese Adults, United States
Why?
Data Source: NHANES 20202010
Who?
What?
Syndemics
Prevention Network
Questioning the Character of Public Health Work
PUBLIC HEALTH WORK
InnovativeHealth
Ventures
SYSTEMS THINKING & MODELING (understanding change)
• What causes population health problems?
• How are efforts to protect the public’s health organized?
• How and when do health systems change (or resist change)?
PUBLIC HEALTH(setting direction)
What are health leaderstrying to accomplish?
SOCIAL NAVIGATION(governing movement)
Directing Change
Charting Progress
• Who does the work?• By what means?• According to whose values?
• How are conditions changing?• In which directions?
Syndemics
Prevention Network
EXTRAS
Syndemics
Prevention Network
Potential Users and Uses of Health SD Simulation Models
• Planners/Evaluators/Media: Chart Progress Toward Goals– Define a “status quo” future– Define alternative futures based on policy scenarios– Define types of information to be routinely collected – Track and interpret trajectories of change– Estimate how strong interventions must be to make a difference
• Researchers: Better Measurement and New Knowledge– Integrate diverse data sources into a single analytic environment – Infer properties of unmeasured or poorly measured parameters– Analyze historical drivers of change– Locate areas of uncertainty to be addressed in new research
• Policy Makers: Convene Multistakeholder Action Labs– Understand how a dynamically complex system functions– Discover short- and long-term consequences of alternative policies– Prepare for difficult patterns of change (e.g., worse-before-better)– Consider the cost effectiveness of alternative policies– Explore ways of combining and aligning policies for better results– Increase policy-makers’ motivation to act differently
• Others…
Syndemics
Prevention Network
Possible Roles for System Dynamics in Public HealthSD is especially well-suited for studying…
• Individual diseases and risk factorsExamining momentum and setting justifiable goals
• Life course dynamics Following health trajectories across life stages
• Mutually reinforcing afflictions (syndemics)Exploring interactions among related afflictions, adverse living conditions, and the public’s capacity to address them both
• Capacities of the health protection system Understanding how ambitious health ventures may be configured without overwhelming/depleting capacity--perhaps even strengthening it
• Value trade-offs Analyzing phenomena like the imbalance of upstream-downstream effort, growth of the uninsured, rising costs, declining quality, entrenched inequalities
• Organizational management Linking balanced scorecards to a dynamic understanding of processes
• Group model building and scenario planningBringing more structure, evidence, and insight to public dialogue and judgment
Syndemics
Prevention Network
Steps for Developing Dynamic Policy Models
Enact PoliciesBuild power and organize actors to
establish chosen policies
Enact PoliciesBuild power and organize actors to
establish chosen policies
Choose AmongPlausible Futures
Discuss values and consider trade-offs
Choose AmongPlausible Futures
Discuss values and consider trade-offs
Learn About Policy Consequences
Test proposed policies, searching for ones that best
govern change
Learn About Policy Consequences
Test proposed policies, searching for ones that best
govern change
Run Simulation Experiments
Compare model’s behavior to expectations and/or data to
build confidence in the model
Run Simulation Experiments
Compare model’s behavior to expectations and/or data to
build confidence in the model
Convert the Map Into a Simulation Model
Formally quantify the hypothesis using allavailable evidence
Convert the Map Into a Simulation Model
Formally quantify the hypothesis using allavailable evidence
Create a Dynamic Hypothesis
Identify and map the main causal forces that
create the problem
Create a Dynamic Hypothesis
Identify and map the main causal forces that
create the problem
Identify a Persistent Problem
Graph its behavior over time
Identify a Persistent Problem
Graph its behavior over time