Enhancing implementation science by applying best principles of systems science
- 2.4k Downloads
Implementation science holds promise for better ensuring that research is translated into evidence-based policy and practice, but interventions often fail or even worsen the problems they are intended to solve due to a lack of understanding of real world structures and dynamic complexity. While systems science alone cannot possibly solve the major challenges in public health, systems-based approaches may contribute to changing the language and methods for conceptualising and acting within complex systems. The overarching goal of this paper is to improve the modelling used in dissemination and implementation research by applying best principles of systems science.
Best principles, as distinct from the more customary term ‘best practices’, are used to underscore the need to extract the core issues from the context in which they are embedded in order to better ensure that they are transferable across settings. Toward meaningfully grappling with the complex and challenging problems faced in adopting and integrating evidence-based health interventions and changing practice patterns within specific settings, we propose and illustrate four best principles derived from our systems science experience: (1) model the problem, not the system; (2) pay attention to what is important, not just what is quantifiable; (3) leverage the utility of models as boundary objects; and (4) adopt a portfolio approach to model building. To improve our mental models of the real world, system scientists have created methodologies such as system dynamics, agent-based modelling, geographic information science and social network simulation. To understand dynamic complexity, we need the ability to simulate. Otherwise, our understanding will be limited. The practice of dynamic systems modelling, as discussed herein, is the art and science of linking system structure to behaviour for the purpose of changing structure to improve behaviour. A useful computer model creates a knowledge repository and a virtual library for internally consistent exploration of alternative assumptions.
Among the benefits of systems modelling are iterative practice, participatory potential and possibility thinking. We trust that the best principles proposed here will resonate with implementation scientists; applying them to the modelling process may abet the translation of research into effective policy and practice.
KeywordsBest principles Complexity Context Implementation science Modelling Health equity Oral health Primary care Screening at chairside Systems science
This review is grounded in the ongoing experiences of the authors to devise and implement interventions to promote health equity, including for older adults. Because the aforementioned interventions are both multilevel and dynamic, the scientific approaches employed evolved from utilising ecological models for thinking through pathways whereby determinants at the societal, community and interpersonal levels affect population and individual health and well-being [1, 2, 3, 4], to embracing a portfolio of systems science models that usefully inform related research, practice, policy and education initiatives [5, 6, 7].
Forrester, the founder of system dynamics, famously explained that a manager’s verbal description of a corporate organisation constitutes a model . Such mental models of corporations are used by managers to deal with problems that arise on a daily basis. They are not, however, the real corporation. Rather, they substitute in our thinking for the real organisation. Sterman, a leading systems scientist modeller and extraordinary communicator, attributes the lack of learning effectively in a world of dynamic complexity to poor inquiry skills. He argues, “We do not generate alternative explanations or control for confounding variables. Our judgments are strongly affected by the frame in which the information is presented, even when the objective information is unchanged. We suffer from overconfidence in our judgments (underestimating uncertainty), wishful thinking (assessing desired outcomes as more likely than undesired outcomes), and confirmation bias (seeking evidence consistent with our preconceptions)” (, p. 510).
A complex (adaptive) system has been usefully defined as a system comprised of a large number of entities that display a high level of interactivity that is largely nonlinear, containing demonstrable feedback loops [10, 11]. The term systems science is used to refer to the ‘big picture’ of problem solving, where the problem space is conceptualised as a system of interrelated component parts . Both the coherent whole of the system and the relationships among the component parts are critical to the system, as they give rise to emergence, meaning much coming from little . Note that emergence occurs when even a relatively simple system generates unexpected amounts of complexity, which cannot be understood without the ability to create a model . There are a number of other basic observations that have been made through the examination of complex systems, primarily through the use of computer simulation and the mathematics of nonlinearity, including self-organisation, meaning insensitive to large disturbances  and incompressibility, meaning any reduction in complexity will result in the loss of system aspects . The overarching point is that rather than focusing on the parts of a system and how they function, one must focus on the interactions between these parts, and how these relationships determine the identity not only of the parts, but of the whole system .
Likewise, dissemination and implementation research places an emphasis on studying issues in context [3, 16, 17]. In his seminal article on diffusion, dissemination and implementation, Lomas explained, “Implementation … is dependent on a complex framework of sanctions and incentives, reinforced by monitoring and adjustment, and often adapted to fit differing environments at more local levels” (, p. 227). Thus, the congruence of an implementation science approach with a systems science approach is both intuitive and pragmatic. After first-hand engagement in conducting an implementation science pilot study [19, 20], however, the use of systems science modelling to strengthen the dissemination and implementation evidence base became a tangible next step rather than a future direction for the field .
Previous researchers have contended systems thinking may usefully advance implementation science. Indeed, Glasgow and Chambers  argued that implementation researchers would profit from embracing an interrelated systems perspective rather than a mechanistic, determinism approach to science. Further, Holmes et al.  sought to draw attention to certain implications inherent in adopting a systems view for dissemination and implementation research, especially with regard to causation and leverage points for change in a complex system. Recently, Burke et al.  presented case examples of three systems science methods, namely system dynamics, agent-based modelling and network analysis, to illustrate how each method may be used to address dissemination and implementation challenges. Finally, Valente conducted a review of network interventions without specifically relating them to implementation science, yet concluded that the choice of intervention depends, in part, on the social context of the program , in concert with the systems perspective that context is critical .
While complex systems science alone cannot possibly solve the major challenges in public health, it has been argued that systems-based approaches may contribute to changing the language and methods for conceptualising and acting within complex systems . Moreover, it may eventually improve the modelling used in dissemination and implementation research. Toward that end, we thought to share best principles of systems science that we have successfully applied in our own studies toward enhancing implementation science. Best principles, as distinct from the more customary term best practices, are used to underscore the need to extract the core issues from the context in which they are embedded in order to better ensure that they are transferable across settings . For a full treatment of the principles, meaning fundamental truths, of systems science, see the recent text by Mobus and Kalton .
The Modelling Process
The problem we were attempting to solve in our pilot study was to improve primary care screening and care coordination at chairside, meaning in a dental setting rather than a medical or other setting . While we had both championed and been involved in previous initiatives that integrated oral health and primary care [29, 30, 31, 32], our idea was to support dental hygienists in practicing to the full extent of their training so that they might effectively implement evidence-based guidelines for tobacco use, hypertension and diabetes screening, and nutrition counselling in dental settings . We are principally focused on advancing health equity and ensuring that population groups who lack oral health and primary care are linked to accessible providers and care settings in their own communities, whenever possible [7, 30].
Next, we propose and illustrate four best principles derived from our ongoing systems science research and scholarship that may guide, and perhaps even motivate, implementation scientists in their own studies and thinking. The overarching theme of these best principles involves meaningfully informing the modelling process. It is our belief that this aspect of dissemination and implementation research demands concerted attention in order to meaningfully grapple with the complex and challenging problems faced in adopting and integrating evidence-based health interventions and changing practice patterns within specific settings .
Best Principle #1: Model the problem, not the system
Sterman rightly deserves credit for driving home the importance of modelling the problem, not the system . Accordingly, we began our aforementioned pilot study by conducting formative research about the views of dental providers (both dental hygienists and dentists) on primary care coordination at chairside . Findings were that both the dental hygienists and dentists interviewed as part of this research failed to use evidence-based guidelines to screen their patients for primary care-sensitive conditions such as hypertension and diabetes . Nonetheless, all of the participating dental hygienists and dentists reported using electronic devices at chairside to obtain web-based health information in caring for their patients . Hence, we worked collaboratively to develop a clinical decision support system for use by dental hygienists to support them in providing patient care at the level of their full scope of practice [19, 33].
In subsequent research, we reframed the locus of concern around health equity more broadly, requiring us to reconsider how an individual’s health status reflects a broader distribution of social and health disparities that vary by population subgroups. An orientation toward health equity warrants a broader model conceptualisation than health per se .
Attempts to model the system rather than the problem are bound to lead to confusion and futility . Our training and experience in systems science directed us away from trying to design an integrated system of oral and primary care and focused our attention instead on supporting dental hygienists to adhere to evidence-based tobacco use, hypertension and diabetes screening, and nutrition counselling referral guidelines [19, 20, 33]. Formative research and interdisciplinary collaboration are invaluable in steering implementation scientists toward modelling the problem, not the system.
Best Principle #2: Pay attention to what is important, not just what is quantifiable
Meadows was a rigorous systems scientist who inspired her colleagues and students to pay attention to what is important – be it justice, democracy, security, freedom, truth, or love – even if it cannot be precisely defined or measured . Unfortunately, despite the critical importance of qualitative information, certain researchers restrict the constructs and variables in their models to those for which numerical data are available, and include only those parameters that can be estimated statistically . Yet, in a comprehensive article on collecting and analysing qualitative data for system dynamics , Luna-Reyes and Andersen argue convincingly that qualitative data and their analysis also have a central role to play at all stages of the modelling process. Using strategies such as theirs, qualitative statements can be used to derive causal relationships.
Because dissemination and implementation studies are based on the mechanisms through which health information, interventions and evidence-based clinical practices are adopted in public health, community and healthcare service use in a variety of settings, a broad range of methodological approaches are employed . These include both traditional designs, such as randomised controlled trials, and newer approaches such as hybrid effectiveness-implementation designs [40, 41]. While mixed methods approaches are endorsed in implementation science, there is a need for greater attention to connectedness across program levels and components .
We are at the point in our implementation science study of primary care coordination by dental hygienists at chairside where we need to create a causal map (also known as causal loop diagram) to provide a systematic way to develop dynamic hypotheses and identify important feedback loops . In a causal map, it is possible to ascribe certain variables to specific scales, e.g. community, interpersonal and individual. Because systems science models are not limited to constructs that are precisely defined or measured, deep thinking and multiple perspectives may help guide implementation scientists to pay attention to what is important, not just what is quantifiable.
Best Principle #3: Leverage the utility of models as boundary objects
According to Black, a boundary object is “a representation—perhaps a diagram, sketch, sparse text, or prototype—that helps individuals collaborate effectively across some boundary, often a difference in knowledge, training, or objective” (, p. 76). For research teams such as ours, whose members possess expertise in diverse domains, boundary objects are useful for coordinating knowledge and objectives and for developing a shared vocabulary about the problem to be solved collaboratively .
This graphic proved to be both intuitive and accessible to our interdisciplinary team members, so much so that we have created project-specific models for a series of papers [19, 20, 33]. We now consider our CFIR model to be a boundary object that facilitates team collaboration.
Note that, from a modelling perspective, a boundary object is “a socially constructed artefact for building trust and agreement” (, p. 4, citing ). For boundary objects to be useful, they must be modifiable and readily perceptible representations that embody the dependencies among resources and goals of team members . While boundary objects represent local knowledge, they may be shared across networks and thus play a significant role in creating synergies which in turn sustain local initiatives . Developed models used as boundary objects may benefit implementation scientists through building trust and agreement that represent local knowledge.
Best Principle #4: Adopt a portfolio approach to model building
As we alluded to at the outset of this paper, our research team led by the authors – an implementation scientist (MEN) and a systems scientist (SSM) – has developed a portfolio of conceptual, statistical, spatial and simulation models that utilise the multiple information streams associated with our research projects . A chief advantage of the portfolio approach in a collaborative research context is that it provides multiple entry points and checkpoints to the modelling process, facilitating input from different team members . A further benefit is that team members often work in parallel to develop separate but related models in diverse ways, exploring the simulated consequences of alternative assumptions .
For instance, in our ongoing project, Integrating Social and Systems Science Approaches to Promote Oral Health Equity, our modelling team has gained important insights by adopting a portfolio approach that incorporates different methods of systems science, including system dynamics, agent-based modelling, geographic information science and social network simulation, in models that help to explore challenges to realising oral health equity for older adults [6, 35]. This portfolio approach to systems science modelling enables our research team to interpret and triangulate between different scenarios at distinct geographic and temporal scales. An inventory of the simulation models in our portfolio that highlights their links to other models in the portfolio is provided in Additional file 1.
In essence, then, the construction of a portfolio of models confers flexibility to the modelling process and is especially conducive to collaboration, allowing for multiple opportunities for input and adjustment of models by different members of the research team. Further, the portfolio approach leverages the iterative nature of the modelling process and encourages exploration with ‘flawed’ models rather than aiming for perfection with ‘kitchen sink’ models. Implementation scientists may profit from adopting a portfolio approach to model building that confers flexibility and is conducive to collaboration.
Because as humans we can only process a limited amount of information in our heads as ‘thought experiments’, we need to simulate computer models to transcend our mental models. In short, to understand dynamic complexity, we need the ability to simulate. Otherwise, our understanding will be limited.
Modelling, then, is the art and science of linking system structure to behaviour for the purpose of changing structure to improve behaviour. A useful computer model creates a knowledge repository and a virtual library for internally consistent exploration of alternative assumptions. Among the benefits of systems modelling are iterative practice, participatory potential and possibility thinking.
Summary of best principles from systems science for informing the modelling process, recommendations for action by implementation scientists and contributing thought leaders and key references
Thought leader [Reference]
1. Model the problem, not the system
Conduct formative research; construct models collaboratively in interdisciplinary teams
2. Pay attention to what is important, not just what is quantifiable
Use qualitative data to derive causal relationships; be guided by deep thinking and multiple perspectives
3. Leverage the utility of models as boundary objects
Create modifiable and readily perceptible representations of models; build trust and agreement by representing local knowledge
4. Adopt a portfolio approach to model building
Work in parallel to develop separate but related models in diverse ways; encourage exploration with ‘flawed’ models rather than aiming for perfection
As Sterman cautions us, “What prevents us from overcoming policy resistance is not a lack of resources, technical knowledge, or a genuine commitment to change. What thwarts us is our lack of a meaningful systems thinking capability” (, p. 513).
The authors were supported in the research, analysis and writing of this paper by the National Center for Advancing Translational Sciences of the US National Institutes of Health for the project entitled, Primary Care Screening by Dental Hygienists at Chairside: Developing and Evaluating an Electronic Tool (grant UL1TR000038) and by the National Institute for Dental and Craniofacial Research and the Office of Behavioral and Social Sciences Research of the US National Institutes of Health for the project entitled, Integrating Social and Systems Science Approaches to Promote Oral Health Equity (grant R01-DE023072).
MEN conceived of the study, participated in its design, wrote the first draft, and contributed to the conceptualisation of the figures. SSM participated in the design of the study, provided substantive edits to the draft and created the figures. Both authors read and approved the final manuscript.
The authors declare that they have no competing interests.
Ethics approval and consent to participate
All Columbia University, New York University, and University at Buffalo institutional review board and Health Insurance Portability and Accountability Act safeguards were followed.
- 8.Forrester JW. Industrial Dynamics. Waltham: Pegasus Communications; 1961.Google Scholar
- 13.Holland JH. Emergence: From Chaos to Order. New York: Basic Books; 1998.Google Scholar
- 14.Auyang SY. Foundations of Complex-System Theories in Economics, Evolutionary Biology, and Statistical Physics. Cambridge: Cambridge University Press; 1999.Google Scholar
- 15.Cilliers P. Complexity and Postmodernism: Understanding Complex Systems. London: Routledge; 1998.Google Scholar
- 16.Biglan A. Changing Cultural Practices: A Contextualist Framework for Intervention Research. Reno: Context Press; 1995.Google Scholar
- 19.Russell SL, Greenblatt AP, Gomes D, Birenz S, Golembeski CA, Shelley D, McGuirk M, Eisenberg E, Northridge ME. Toward implementing primary care at chairside: developing a clinical decision support system for dental hygienists. J Evid Based Dent Pract. 2015;15(4):145–51.CrossRefPubMedPubMedCentralGoogle Scholar
- 20.Northridge ME, Birenz S, Gomes G, Golembeski CA, Greenblatt AP, Shelley D, Russell SL. Views of dental providers on primary care coordination at chairside. J Dent Hyg. 2016;90(3):195–205.Google Scholar
- 23.Holmes BH, Finegood DT, Riley BL, Best A. Systems thinking in dissemination and implementation research. In: Brownson RC, Colditz GA, Proctor EK, editors. Dissemination and Implementation Research in Health: Translating Science to Practice. New York: Oxford University Press; 2012. p. 175–91.CrossRefGoogle Scholar
- 27.Sclar ED, Northridge ME, Karpel EM. Promoting interdisciplinary curricula and training in transportation, land use, physical activity, and health. In: Does the Built Environment Influence Physical Activity? Examining the Evidence. Transportation Research Board Special Report 282. Washington, DC: Transportation Research Board; 2005.Google Scholar
- 34.Sterman JD. Business Dynamics: Systems Thinking and Modeling for a Complex World. New York: The McGraw-Hill Companies, Inc.; 2000.Google Scholar
- 35.Metcalf SS, Northridge ME. Engaging in systems science to promote health equity. SAGE Research Methods Case. (in press)Google Scholar
- 36.Meadows DH. Thinking in Systems: A Primer. White River Junction: Chelsea Green Publishing Co.; 2008.Google Scholar
- 39.Brownson RC, Colditz GA, Proctor EK, editors. Dissemination and Implementation Research in Health: Translating Science to Practice. New York: Oxford University Press; 2012.Google Scholar
- 42.Metcalf SS, Kum SS. System dynamics. In: Richardson D, Castree N, Goodchild MF, Kobayashi A, Liu W, Marston RA, editors. International Encyclopedia of Geography: People, the Earth, Environment, and Technology. Hoboken: Wiley-Blackwell and the Association of American Geographers; 2016.Google Scholar
- 44.Kum SS, Wang H, Jin Z, Xu W, Mark J, Northridge ME, Kunzel C, Marshall SE, Metcalf SS. Boundary objects for group model building to explore oral health equity. Cambridge: 33rd International Conference of the System Dynamics Society; 2015. http://www.systemdynamics.org/conferences/2015/papers/P1302.pdf. Accessed 3 Mar 2016.Google Scholar
- 47.Zagonel AA. Model conceptualization in group model building: a review of the literature exploring the tension between representing reality and negotiating a social order. Proceedings of the 20th International System Dynamics Conference. Palermo: System Dynamics Society; 2002. http://www.systemdynamics.org/conferences/2002/proceed/papers/Zagonel1.pdf. Accessed 3 Mar 2016.Google Scholar
- 49.Nyella E, Nguyen T, Braa J. Collaborative knowledge making and sharing across sites: the role of boundary objects. Mediterranean Conference on Information Systems (MCIS) 2010 Proceedings. Paper 64. http://aisel.aisnet.org/mcis2010/64. Accessed 3 Mar 2016.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.