Ending the HIV epidemic in Chicago: Evidence from high-fidelity local agent-based model

Wouter Vermeer, PhD
Northwestern University

Nanette Benbow, M.A.S.
Northwestern University

ABSTRACT:
Agent-based models have enormous potential for modeling complex social phenomena. The ability to model complex individual-level social dynamics and project system-wide behaviors can help inform decision makers in their effort to curb a phenomenon like the spread of HIV. To produce actionable results, models need to accurately capture the dynamics and behaviors observed by decision makers. As such, models aimed at supporting decision making need to be tailored to the local context by incorporating behaviors based on local data. In this presentation we describe our model for HIV-spread in Chicago and highlight how we used Chicago-level data and input from local public health experts to inform this model. We will illustrate how our model can be used to perform predictive scenario analysis and discuss how these results can be used to inform decision makers as they plan their HIV care and prevention strategies to end the HIV epidemic.

SYSTEMS SCIENCE AND POPULATION HEALTH

Sandro Galea, M.D., Ph.D.
Boston University

ABSTRACT:
Population health is the product of complex systems. It is characterized by discontinuities, non-linearities, reciprocity, and emergent properties. This requires clarity about our conceptualization of population health and the approaches we take towards answering population health science questions. This presentation will discuss the fundamental principles of population health science and how it intersects with complex systems perspectives, using examples from ongoing modeling work.

Evaluating complex interventions: Confronting and guiding (versus ignoring and suppressing) heterogeneity and adaptation

Brian S. Mittman, Ph.D.
Kaiser Permanente

ABSTRACT:
Implementation strategies and many of the clinical and health service delivery interventions they aim to implement are characterized by multiple components targeting multiple behaviors and levels and are often characterized by extreme heterogeneity and adaptability.  Although researchers often attempt to standardize and achieve fidelity to highly-specified manualized intervention protocols, the required actions to suppress adaptation and maximize internal validity often lead to reduced effectiveness:  adaptation to local conditions often increases intervention effectiveness relative to implementation of a fixed version of an intervention across heterogeneous settings.  This presentation introduces the new PCORI Methodology Committee Standards for Complex Interventions and discusses their role in research to (a) study and guide rather than suppress or ignore adaptation, achieving internal validity through adherence to an adaptation algorithm and through fidelity to function rather than form, and to (b) develop empirical evidence, insights and guidance for policy and practice decision makers who are charged with adapting and managing complex interventions rather than simply selecting and deploying simple, fixed interventions.

Prediction vs. explanation in statistical model building

Otto Koppius, Ph.D.
Erasmus University

ABSTRACT:
Predictive modeling, where one tries to predict for instance the outcome of a particular process or the occurrence of a certain event, is common in many research areas. In medical research, predictive models (often called prognostic models) are for instance used to predict patient outcomes or the effect of certain treatments. A common practice is using the best explanatory statistical model for prognostic purposes, i.e. the model with significant coefficients for the independent variables. While common, this is incorrect for prognostic purposes, as the best explanatory model is almost always different from the best prognostic model for a number of reasons, which I will describe in this talk. Furthermore, the process of building a predictive model is fundamentally different from building an explanatory model, as differences occur in every step of the modeling process. Moreover, predictive models have additional roles to play alongside explanatory models in theory building and theory testing, such as new theory generation, measurement development, comparison of competing theories, improvement of the conceptual structure of existing models, relevance assessment, and assessment of the predictability of empirical phenomena. I will illustrate the differences between the modeling approaches with examples from the literature on adoption of new health technologies and from diffusion over networks.

Testing multitheoretical, mutlilevel hypotheses about networks

Noshir Contractor, Ph.D.
Northwestern University

ABSTRACT:
Networks are agile and are constantly adapting as new links are added and others dropped. We review the theoretical mechanisms that have been used to explain emergence of the networks. We then offer an analytic framework to specify and statistically test simultaneously multilevel, multitheoretical hypotheses about the structural tendencies of networks. Specifically, we focus on Exponential Random Graph models for cross-sectional networks, stochastic actor-oriented models for longitudinal panel network data, and relational event modeling for continuous (time stamped) data. We offer empirical examples to illustrate the capabilities of this framework.

Agent-based modeling applications in healthcare and infectious diseases

Charles Macal, Ph.D.
Argonne National Laboratory

ABSTRACT:
Agent-based modeling (ABM) is an approach to modeling systems comprised of autonomous, interacting agents. ABM applications are growing rapidly in many areas including healthcare: for modeling populations of heterogeneous agents, their behaviors and interactions. Applications range from modeling the onset of disease in future populations, to the diffusion of health resource information throughout the community, to the spread of cancer in the body, to predicting the extent of possible pandemics, among many others. Agent-based models often exhibit emergent, adaptive, behavior, and self-organization. This seminar briefly introduces ABM, discusses when it is appropriate to apply its unique capabilities, and discusses some practical applications related to healthcare.

The impact of the propagation mechanism on system wide dynamics

Wouter Vermeer, Ph.D.
Northwestern University

ABSTRACT:
Often systems exhibit behavior which is difficult to predict and steer, as interactions on the micro level (between actors within the system) results in propagation of behavior which can cause unforeseen dynamics. Understanding the effects of propagation is crucial in order to understand how the state and behavior of a system will change and what the effect of interventions will be. The structure of interactions, the network structure, has been identified in literature as a prime driver of propagation. By focusing on the network structure, however, the impact of the mechanism by which propagation takes place has been pushed to the background. In this presentation I argue that the mechanism plays a crucial role in determining how propagation dynamics scale, and is thus critical when effectively intervening in a system. While various mechanisms exist in literature there is little coherence and overlap in their use, therefore I put forward a generic Agent-Based framework which allows for capturing the propagation mechanism in more detail and in a structured way. I show that using such a framework not only results in a more detailed and methodologically stronger model of propagation, but also that it is a prerequisite for effective interventions into propagation.