The conversion of a complex adaptive systems modeling sceptic

During the past few years complex adaptive systems theory has suddenly become very popular in the health systems research field. Somehow those crazy obesity (click through for a nice interactive version or see image above) and tobacco control causal loop diagrams resonated with researchers in the field. Before we knew what was going on the Alliance for Health Policy and Systems Research was publishing reports on systems thinking and the National Institutes of Health were holding conferences on complex systems and health disparities.

At some level, complex adaptive systems (CAS) are easy to understand and appreciate. As Josh Epstein likes to say we all constitute models in our head: "Anyone who ventures a projection, or imagines how a social dynamic--an epidemic, war, or migration--would unfold is running some model."

Yet it's tougher to see how CAS modeling can really contribute to health systems analysis in low- and middle-income countries, where data are frequently poor and incomplete. What's more, my first impressions of agent based models, which run on simple rules governing actions and interactions of autonomous agents, was that they seem to offer somewhat trite insights.

I have to admit that -- while the notions of tipping points, and emergent behavior and path dependence all made absolute sense to me, and seem like valuable lines of enquiry for more qualitative research -- until recently I was something of a complex systems modeling sceptic. I thought it was fine for those people who work on disease transmission and outbreaks, but not cut out for exploring health worker policy issues, or health financing.

However, a couple of recent events, notably the FHS workshop and the NIH conference mentioned above, have changed my perspective. What has driven my change in thinking?

1. Modeling CAS can be a little scary if not explained well! Previously differences between agent based modeling and systems dynamics models, and how one would go about setting each of these up, were simply not clear to me. As a consequence, they remained remote and impenetrable.
2. CAS modeling is not a solo endeavor, and I don't really need to be a modeler to participate. It seems that CAS modeling is best done in multi-disciplinary teams who collectively explore the nature of relationships between the different variables, bringing different perspectives to bear. This kind of collaborative process can include researchers, practitioners and/or policy makers who are best placed to understand what kind of interventions may be feasible, and sometimes community members, as well as the modelers. Indeed participatory modeling seems to be quite a major force within the CAS field.
3. The need to apply complex adaptive systems models to health systems is increasingly evident. This is a critical point for me, and one that has emerged from our ongoing work. For example, as our FHS colleagues in China struggle to work out how to analyze a health system that seems to be in an almost constant state of change, traditional impact evaluations look increasingly irrelevant. Instead, we need methods that can capture the ripples of health reform throughout a system, identify unforeseen consequences and force us to think more clearly about how context affects an intervention. CAS may help on all of these fronts.

So for those of you who know me well, don't worry, I am not about to become a Vensim or Netlogo whiz… that is too far-fetched! But if you're thinking of a collaborative modeling project on health systems addressing the effects of financial incentives for health workers, or reactions to regulatory reforms for example, then count me in!

By Dr Sara Bennett, FHS CEO, Johns Hopkins Bloomberg School of Public Health