Invited Symposium II: Using System Science Modeling for Prevention Research and Programming

(2-040) Invited Symposium II: Using System Science Modeling for Prevention Research and Programming,

Bayview A

Chair: Elizabeth M. Ginexi, PhD, Program Director, Tobacco Control Research Branch, Behavioral Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute

Discussant: Ty A. Ridenour, PhD, Research Associate Professor, University of Pittsburgh

Presenters: Kristen Hassmiller Lich, PhD, University of North Carolina, Chapel Hill, Georgiy Bobashev, PhD, RTI International, Diane Orenstein, PhD, Centers for Disease Control and Prevention

This paper symposium will bring together a distinguished panel of experts who will describe their use of systems science simulation models for public health research and programming.  Systems science approaches are applicable to prevention science because of their capacity to model empirically the complex and dynamic factors at play. These tools are arguably similar to more traditional methods such as case studies and surveys in their capacity to study individual components of systems, but may be superior in their capacity to reveal how multiple components interrelate. Dynamical system science models can help reveal more holistically how risk and protective factors at multiple levels of influence unfold and influence each other in real world contexts over time and how prevention programs and policies can best be leveraged. Each speaker will present an example of a systems science simulation modeling approach to illustrate how prevention science and programming could benefit from the application of simulation modeling. Following the presentations there will be a facilitated interactive discussion between the panel experts, the discussant and the audience centered on creative applications of system science tools for prevention research and programming. 

Using a Dynamic Computational Model to Support Identification of Policy and Research Priorities in the context of Substantial Uncertainty

Kristen Hassmiller Lich, PhD, University of North Carolina, Chapel Hill

The Veterans Affairs (VA) Stroke Quality Enhancement Research Initiative (QUERI) is a national research program that seeks to reduce the risk and burden of stroke and to foster system, provider, and patient processes to improve stroke outcomes among Veterans. In this presentation, we present the system dynamics simulation model developed to simulate and compare 15 stakeholder-generated intervention scenarios under consideration over the short (5-year) and long-term (20 year) planning horizons. Intervention scenarios include broad and targeted primary prevention, secondary prevention, acute care, and post-stroke rehabilitation. We also introduce our approach to conducting rigorous sensitivity analysis, calibration, and uncertainty analysis with the model to facilitate Stroke QUERI strategic planning around evidence translation and research priorities. Despite substantial uncertainty in many model input parameters, this analysis allowed identification of robust policy recommendations as well as specific targets for future research. This work illustrates how broad simulation models can guide learning early in their life cycle (while many data inputs are quite uncertain) and offers a pathway to processing and presenting complex simulation data in support of robust conclusions.

Agent based models: Relevance to health and policy areas

Georgiy Bobashev, PhD, RTI International

Human behavior is dynamic, which means that it changes and adapts. Health sciences, however, often consider linear and static association-based models. Such relationships could be measured from survey data but they don’t capture causality. When applying interventions or policy changes the assumption of causal relationship is critical for success. Agent-based models (ABMs) allow one to consider causality directly through modeling behavior rules, feedbacks, adaptation, and eventually a response to an intervention. Using simulations one can assess how individuals and thus population in general can respond to different implementation scenarios. Agent-based models, however, come at a price. We need to obtain information about the rules and data for behavior parameters. Additionally, model calibration and validation methodology for ABMs has not been rigorously developed. This sometimes creates confusion about the utility of ABMs. I am going to clarify the difference between theoretical ABMs that illustrate the concept and do not need rigorous underlying data and practical ABMs that are aimed to describe a specific phenomenon and identify actionable steps. I will give examples of model applications to prevention of infectious disease spread in a community and to the evaluation of a substance use policy. I will emphasize the role ethnographic research can play in the understanding and modeling of human behavior.

Prevention Impacts Simulation Model: Bending the Curve

Diane Orenstein, PhD, Centers for Disease Control and Prevention

What are the most effective and economical strategies for reducing chronic illness? Both practitioners and policy makers commonly pose this question when weighing intervention options and costs. Practical answers are often hard to obtain, in part, because most analytic tools have narrow boundaries, short time horizons, and incomplete causal structures that are unable to capture important aspects of chronic disease dynamics. To support more effective collective decision making, the Centers for Disease Control and Prevention (CDC), with additional collaboration the Office of Behavioral and Social Science Research (OBSSR) at the NIH, and the National Heart, Lung and Blood Institute (NHLBI) have created the Prevention Impacts Simulation Model (PRISM): a health policy simulator that brings greater structure, evidence, and creativity to the challenge of reducing the burden of chronic disease. PRISM’s scope encompasses cardiovascular disease (CVD), diabetes, obesity, blood pressure, cholesterol, smoking, secondhand smoke exposure, physical activity, diet, air pollution, and emotional distress. PRISM also represents 34 potential policy interventions that affect health behaviors, environmental exposures, and disease progression through a range of channels such as availability and access, price, promotion, consumption, regulation, social support, and health service utilization.