The term systems science is used here to refer to bringing to problem solving a perspective in which the problem space is conceptualized as a system of interrelated component parts (i.e., the “big picture”). This term was chosen in lieu of several others that may be synonymous, such as systems thinking or complexity, because some terms are associated with a particular “brand” of thought, and the authors feel that systems science is neutral while also inclusive. The system is viewed as a coherent whole, while the relationships among the components are also recognized and seen as critical to the system, for they give rise to the emergent properties of the system. Emergent properties are those properties that can only be seen at the system level and are not attributes of the individual components themselves (e.g., a flock emerges when a group of birds flies together; it is a property of the system, not of any individual bird). Systems science offers insights into the nature of the whole system that often cannot be gained by studying the component parts in isolation. Moreover, in a systems approach, there is recognition that embedded in the system are feedback loops, stocks and flows, that change over time (i.e., dynamic, nonlinear, complexity of the system).
The advantages of utilizing systems science as a complementary method for addressing complex problems include the fact that nonlinear relationships, the unintended effects of intervening in the system, and time-delayed effects are often missed with traditional reductionist approaches, whereas systems approaches excel at detecting these. The common conceptual orientation that defines a systems approach can be summarized as follows:
a paradigm or perspective that considers connections among different components, plans for the implications of their interaction, and requires transdisciplinary thinking as well as active engagement of those who have a stake in the outcome to govern the course of change.25
Systems science is not a single discipline; rather, it is a linkage of disciplines to bring about problem understanding and solving under the paradigm described above.
Systems science does not refer to a single methodology; rather, it encompasses a wide range of methods and tools (e.g., system dynamics simulation, agent-based modeling, network analysis, Markov modeling, soft-systems analysis, discrete-event modeling). While technology is used to maximize the effectiveness of systems approaches, systems science is not a technology. For an in-depth introduction to this topic, readers are encouraged to view webcasts of the 2007 Symposia Series on Systems Science and Health.47
By embracing systems science, the research community will be better equipped to handle the policy-resistant problems that abound in public health. Policy resistance
refers to the “tendency for interventions to be defeated by the system's response to the intervention itself.” 21
In the last decades of the 20th Century, almost in parallel to the developments that spawned systems biology, the social–ecologic model emerged as a dominant world view in searching for explanations of the broader population-level causes of the very same common, chronic diseases that are the focus of biomedicine today.48–51
Other troubling causes of poor health and shortened life expectancy, such as access to care and disparities and inequality in healthcare delivery, have also been studied. The population, behavioral, and social sciences advanced beyond single discipline and simple causal views toward another valid systems view of understanding health and disease. In this world view, human behavior can be broadly defined as hierarchically organized along levels of complexity, from individual behavior to collective behavioral patterns within groups to higher levels of the clustering of patterns of behavior that are embodied in neighborhoods, worksites, schools, communities, cultural, ethnic, or religious affiliations, to even broader patterns determined by societal norms, financial incentives, and policies. These higher-order levels of factors interact in complex, dynamic, and multifactorial ways to produce the so-called “causes of the causes” of the complex common, chronic diseases.2
In this ecologic perspective, the view of the ultimate “causes of the causes” lies as much in the behavioral–social–ecologic environment as it does in the proximal biological environment evident through reductionist approaches.
The implication of these disparate world views of causation (biomedical and ecologic) calls for a broader integration of the disciplines than has occurred to date. OBSSR's view is that there should be a “macro” integration of the three broad disciplinary domains: the largely biomedical sciences, the largely individual behavioral sciences, and the largely group or population-level sciences of the ecologic world view.
Recently there has been a call for a new integrative vision among the behavioral, social, and public health sciences that might loosely be termed systems socio-behavioral science
, systems medicine
, or, as one author has put it, populomics
This is being called vertical
integration, that is, integration across
rather than within
the three broad domains (i.e., the biomedical; the individual behavioral [intra-individual variation]; and the population [inter-individual or cluster variation] levels) of systems structure.20
The hope is that this type of vertical synthesis across varying levels of analysis will lead to a next generation of science enabling further breakthroughs in the understanding and reduction of the burden and suffering of the major common, chronic diseases that afflict the U.S., other developed nations, and, increasingly, the developing nations. OBSSR's call for systems science is a call for an increasingly global perspective on the interaction, connectivity, and relationships within and across nations. The specific objectives for OBSSR with regard to systems science are:
- To facilitate the development and application of the conceptual frameworks and tools needed for the application of systems methodologies to problems of health and its determinants;
- To promote and support the development of informatics tools to facilitate the collaboration and dissemination of data relevant to the behavioral, population, and social sciences (e.g., longitudinal epigenetic, biomarker, social, and behavioral data related to health);
- To contribute to the development of analytical frameworks, methods, and algorithms capable of integrating, analyzing, and interpreting highly diverse data with varying metrics from research on genomic sequences, molecules, behavior, and social systems;
- To collaborate in the development of the curricula, modules, and materials required to train health scientists in the application of systems science; and
- To encourage the application of systems-organizing principles among stakeholder organizations in behavioral and social sciences research, and to promote the development of systems-organizing expertise among leaders, policymakers, and researchers.
Bringing systems science to bear on public health problems has the potential to explain how small changes at the individual level accumulate at the population level to reveal significant shifts in the absolute causes of disease.2,3
System dynamics modeling and agent-based models are methods that can simulate the complex relationships among the components of a system and emergent behavior—that is, behavior that is observed at the bird's-eye vantage point of the system emerging from the behavior of the individual components of the system (e.g., blood clotting and scab formation emerge at the systems level from the behavior of individual cells). Because of its unique ability to consider simultaneously both the whole system and its individual parts, systems science is capable of producing solutions that take into account a broad range of factors pertinent to the problem under consideration; for instance, genetic-to-environmental–, cellular-to-behavioral–, and biological-to-social–systems approaches have proven extremely valuable when applied to problems identified in a variety of disciplines, including defense,53
and cellular biology.55,56
Systems science shows promise for unlocking the secrets of complex, multidimensional health issues and for transforming this knowledge into effective interventions that can fundamentally change population health.57
An example of applying systems science to public health problems is illustrated by Jones et al.,58
who used system dynamics simulation modeling to explain type 2 diabetes prevalence since 1980 and to predict possible futures through 2050. The conceptual model () divided the U.S. population into those who do not have diabetes (normal glycemic levels); those at high risk for developing type 2 diabetes (i.e., people with prediabetes, divided into diagnosed and undiagnosed); and those who meet diagnostic criteria for type 2 diabetes (diagnosed and undiagnosed, subdivided into with and without medical complications from diabetes). The conceptual model included births (entry into the system); deaths (exit from the system); and individual members' movements among the diagnostic categories over time (stocks and flows), as well as numerous factors contributing to diabetes outcomes (e.g., clinical management of diabetes, self-monitoring, healthy-lifestyle adoption, and medication use).
The relationships among all of these variables were quantified and the model was calibrated and validated in an iterative process using historical data from a variety of sources (e.g., the U.S. Census Bureau, the National Health Interview Survey, the National Health and Nutrition Examination Survey, and the Behavioral Risk Factor Surveillance System).
Simulations were then generated according to a variety of assumptions that were programmed into the model via algorithms. shows the results of the simulated population burden of diabetes (i.e., deaths) under various scenarios where an intervention is introduced that is designed to: (1) improve the clinical management of those diagnosed with diabetes; (2) improve pre-diabetes management; and (3) prevent diabetes (through the prevention of obesity). These three hypothetical scenarios are compared to “baseline,” a predictive model in which the status quo of diabetes clinical practices and prevention activities is maintained at baseline levels.
Model output for three intervention scenarios compared with the baseline scenario for diabetes complication-related deaths
The following outcomes were predicted under each of the three scenarios:
- The improved clinical management of diabetes leads to short-term improvements in diabetes control, complications, and associated deaths. However, following these improvements in the first few years, there is a rapid rise in complication deaths. Improvements in complications are rapidly overtaken by the growth in diabetes prevalence because nothing has been done to reduce diabetes onset.
- Efforts to manage persons with prediabetes would lead to reductions in the onset of diabetes initially, and ultimately would reduce deaths from diabetes complications. But without prediabetes prevention efforts, the amount of reduction in deaths is less than optimal.
- Finally, the primary prevention of diabetes shows the most drastic and lasting reductions in deaths.
However, even this powerful step alone (i.e., reducing rates of obesity without concurrent changes in prediabetes management or clinical diabetes management) would not reduce the overall burden of diabetes in terms of both the number of unhealthy days (not pictured) and the number of deaths due to diabetes right away (). In fact, the number of deaths attributable to diabetes would actually rise through at least the year 2020, although during subsequent decades, a significant decrease in diabetes prevalence and deaths would occur. Thus, the time perspective is vital to determining the value of a strategy—that is, disease management works in the short term, but primary prevention is more effective in the long term. This example illustrates the potential of systems science to inform healthcare and policy decisions to improve population health.
In another example of adopting a systems approach to improving the understanding of a public health problem, Levy and colleagues developed SimSmoke,59
a simulation model for guiding policy to make a population impact on reducing smoking prevalence. SimSmoke uses historical and current data to model the multiple sources and complex interrelationships that determine tobacco-use prevalence and its health effects. A discrete-time dynamic model was developed that simulated smoking prevalence and tobacco-related deaths over a 40-year period. The model employed a first-order Markov process that modeled population growth and age-based rates of tobacco initiation, cessation, and relapse. This model simulated the impact of five policy-level interventions on smoking prevalence: taxes, clean indoor-air laws, strategies to reduce youth access to cigarettes, strategies to promote cessation treatments, and mass-media policies. Researchers used empirical and predicted data for the effects of each of these areas on model parameters. SimSmoke showed the relative contributions made by a variety of different policy interventions (i.e., increasing cigarette prices, introducing smoking bans, introducing media campaigns to encourage cessation and prevention, implementing additional restrictions on youth access to tobacco, and introducing proactive quitlines) toward the desired outcomes (i.e., reduction in smoking prevalence and reduction in deaths attributable to tobacco). Such models can be used to inform decisions about how best to allocate financial resources and formulate policies to optimize a desired public health impact. The focus is on making an efficient population impact to address a complex societal problem (tobacco-use behavior) with an emphasis on outcomes and on multiple causal pathways, feedback loops, and control-systems dynamics that underlie the way the tobacco industry and the public health constituencies vie for their respective goals.
The above examples illustrate the potential for systems science to radically transform the behavioral, social, and population sciences to a degree similar in magnitude to the transformation that systems biology and bioinformatics are now bringing about in biology. This sentiment is captured in the broad vision for cyber-infrastructure outlined in the Atkins report of the National Science Foundation30
The opportunity is here to create cyberinfrastructure that enables more ubiquitous, comprehensive knowledge environments that become functionally complete for specific research communities in terms of people, data, information, tools, and instruments and that include unprecedented capacity for computational, storage, and communication…. They can serve individuals, teams and organizations in ways that revolutionize what they can do, how they do it, and who participates.