Tuesday, 2 September 2014

Ten things to know about causal loop diagrams (CLDs)

Editor's note: This blog is part of a series of reflections emerging from a workshop on complex adaptive systems research methods held in Baltimore in June 2014.

Publications on causal loop diagrams (CLDs) have increased fourfold since 2007!
Okay, that probably sounds more impressive than it actually is – in reality it means that we've risen from one publication per year to four. Though, this year, that number is likely to jump even further with the recent release of the Alliance for Health Policy and Systems Research new journal supplement on systems thinking for health research.

While the number of publications using CLDs remains negligible, in the grand scheme of things, the appearance of causal loop diagrams on PubMed's radar signals that researchers are increasingly interested in studying complexity and applying new tools and approaches to their research questions.

As discussed at a methods workshop held in Baltimore in June 2014, systems mapping and conceptualization – a key part of the puzzle of intervening in complex systems – is a process of developing diagrams to illustrate system boundaries and the connections between variables.
CLDs are one type of system map, but they are by no means the only one. What makes CLDs particularly useful is that, in addition to mapping a system structure, they help build understanding of connections between a range of variables.

The increased momentum around complexity and tools such as the CLDs begin to raise issues that question the very core of the paradigms in which we generally operate in public health. That creates demand for understanding what CLDs are and when to develop and use them. It also creates an opportunity to drop some of the communication barriers that exist between disciplines and to push the boundaries of our understanding of complex systems.

What CLDs are
  1. Diagrams that make implicit mental models explicit and that identify which components/variables are part of a system of interest and how they are related – including the directionality of the influence, and the feedback that arises.
  2. First developed in engineering, CLDs benefit from a clear description of conventions used for drawing them – a common "language" that can contribute to bridging the gap between disciplines and facilitate team synergies.
  3. A process by which to document our assumptions and biases, and tease out relationships between variables (if it's not clear whether A causes B to grow, then the pathway between the variables contains additional variables that must be unhidden and teased out), as well as brainstorm about variable relationships and influence pathways.
  4. An opportunity to bring together different stakeholders – that might have divergent mental models – in order to facilitate agreement around how and why to intervene in a system, potentially through participatory or group model building.
  5. A way to illustrate relationships stemming from quantitative and qualitative data, as well as to theorize about potential hypothetical relationships, which can be tested through future data collection.
What CLDs are not
  1. NOT the only approach use to conceptualize and map systems
  2. NOT meaningful on their own, without a description of the assumptions underlying its construction and of the context for which it was developed.
  3. NOT always resulting in generalizable findings – only to the extent that the research design and data collection intended and allowed it to be.
  4. NOT the end product – the process can be sometimes more important than the diagram itself.
  5. NOT easy to communicate, particularly to individuals who were not involved in the design of the diagram – however, this is an issue that practice and future research should be able to improve.
The way(s) forward?

Some might argue that conceptual frameworks, decision trees, flow charts are already used for similar purposes. And those same people might therefore question whether our complex world really need another complex diagram, especially if it is difficult to communicate.

To those detractors (well, sceptics maybe), I would argue that there is growing consensus in international development that greater focus is needed on iterative design and flexible implementation approaches. The systems science toolbox can potentially help in this new quest, and CLDs are an approach worth considering. They are just one component of an iterative process – one that we can return to frequently as a system or our knowledge of a system changes.

The purpose that CLDs serve depends on the research questions and intentions, whether it be: building new theories, designing or planning an intervention, identifying variables and relationships for the purpose of building a model (e.g. a system dynamics model), and/or interpreting research findings that are conflicting.

Regardless of the purpose, or the type of diagram for that matter, my advice is to keep it simple and easy to communicate. There is software to help with drawing neat diagrams (e.g. Vensim), however, pen and (multiple sheets) of paper can also do the job. Furthermore, the ideal model building process – and therefore also the ideal CLD development process – should be a team effort. It should also be one that brings together researchers from various fields and perspectives.

By Ligia Paina, FHS researcher, Johns Hopkins Bloomberg School of Public Health