Connecting the dots in Urban Mental Health using Complexity Science
Complexity Science is the language that binds the research of the Centre for Urban Mental Health. But what is it exactly? And how can it help us understand the dynamics of mental health taking shape in an urban environment? Peter Sloot, Professor of Complex Adaptive Systems and board member of the Centre for Urban Mental Health, explains.
To dive right in. What is complexity science?
‘To answer that question, I think we need to define what complex systems are. Looking at it from the bottom up, complex systems are composed of individual sub-systems, which interact with each other in a non-linear way. Think for instance of a city or an ecosystem. A particular aspect of complex systems is that they often exhibit emergent properties, systemic characteristics that spontaneously manifest themselves but do not exist at the sub-system level. Think for instance of water, it does not make sense to ask how wet one molecule of water is, wetness is an emergent property of many interacting water molecules. Or think of our brain. We know neurons are involved in the ability to think, but how much of a thought can you find in a single neuron? A thought is an emergent property of millions or billions of firing neurons. If we take a complex system apart to study it, we might lose the very thing we are looking for, because of these emergent properties that will disappear when breaking the system down into its parts. So yes, the whole is more than and different from the sum of its parts. Complexity science, now, is a computational and mathematical way to get a grip on the dynamics of these complex systems and try to predict and influence their behavior over time.’
Why is that a suitable approach for studying mental health in an urban environment?
‘We understand more and more that there is not a single cause for a mental disorder. For example: there is not a single gene or unfortunate life event, like losing your job, that causes a depressive episode. A better way to understand depression is to view it as a characteristic emerging from a network of symptoms, like sleep problems, rumination, apathy etc, which themselves arise because of an array of factors. All these factors may interact with each other in a non-linear way such that the pathways from cause to effect are obscured.’
Can you give an example of these causes and their interactions in an urban environment?
‘You can think of a range of things. Noise pollution, for example, can cause sleeping problems. Or financial uncertainty, which can cause both stress and sleeping problems. Additionally, this financial uncertainty might coincide with exposure to noise pollution due to poor quality housing. And then these sleeping problems might hamper achievements at work, which in turn might add to financial uncertainty, etcetera. But you can also think of a host of other factors: like food intake influencing the gut microbiome, which can influence mood. And socio-economic capital determining the quality of food intake and so on and so on. If we study any of these causes in isolation, we lose sight of the bigger picture.’
Why do regular research tools fall short when studying complex systems?
‘The science we have developed over the past three centuries is all about reductionism. Think about physics: breaking water down into molecules and molecules into atoms and atoms into quarks. We became phenomenally good at that, but in the process we lost a lot: we neglected the wide array of complex systems surrounding us: ecosystems, the economy, the immune system, mental health all interacting with each other.’
‘In complexity science, we are looking for similarities between different complex systems – for instance the dynamics that cities and ecosystems and the economy have in common. The field of complexity science is looking for universal methods to reason about these complex systems and to reason about intervening in these systems. This is important because of the non-linear interactions between the underlying factors. As we know from some disastrous ecosystem interventions. For example, rabbits were introduced in Australia as they were brought in as pets, by famine migrants some 150 years ago. As they experienced no natural predators, they started to become a plague. So it was decided to intervene and introduce foxes to control the rabbit population. But unfortunately, the foxes liked the slower and more tasty Australian marsupials better. The result was more foxes, more rabbits and a dramatic loss of indigenous fauna; the natural habitat completely disturbed. The latest idea now is to introduce viruses to fight the foxes, but before we do that we should better understand the knock-on effects that such intervention will have. If we do not take these feedback loops and cascading effects into account, we can trigger unintended consequences with our interventions. Looking at it from the positive side: if we do take these interactions between different causal mental health factors into account, we might be able to achieve substantial improvement of mental wellbeing with relatively simple and affordable interventions.’
Thinking about interventions when it comes to mental health, antidepressants and cognitive behavioral therapy come to mind. How do you perceive these when looking at it from a complex systems angle?
‘We could say antidepressants are a form of symptom relief. Like an aspirin is for a headache. It does not take any of the root causes away, but it can relieve the suffering. Therapy is different, in the sense that it generally brings about a learning experience. People might pay more attention to taking care of themselves: sticking to a healthy sleeping pattern, eating better, going outside, working out, etcetera. Or they might learn ways to break patterns of rumination. In that sense, it can influence many of the little components in the complex system that sustain a depression.’
‘Additionally, knowledge of complex systems can help explain why both antidepressants and therapy are only beneficial for a subset of patients. Complex systems show tipping points and path dependence. They can shift quite suddenly from a dynamic balance in state A, let’s say mental wellbeing, to state B, let’s say mental disorder. But getting the system back into state A can follow a very different and longer trajectory. Let’s say lack of sleep was the final push for someone to develop a manic episode, then just restoring the sleep pattern would not necessarily suffice to get the system back into its healthy state. It might require many additional tweaks.’
The field of complexity science is still being developed, but meanwhile researchers at the Centre for Urban Mental Health use the tools of complexity science. How is that possible?
‘We did already develop a whole toolbox of instruments that help us study complex systems [see the case study below]. These can also help us gain understanding of mental health in the city.’
When can we expect to see the first results? And what kind of insights would those be?
‘That is not easy to predict as the Centre for Urban Mental Health in itself is a complex system, with scientists from such a broad range of disciplines: from immunology to psychology, physics, computational and biomedical sciences. Complexity science is the language that connects us in this effort. Having this collective framework helps us structure the research and understand which relations are most important and need to be sorted out more thoroughly. For example, we realize more and more how important inflammation pathways are when it comes to the relationship between the microbiome and mental health. So identifying the best markers for these inflammation pathways can help us understand this relationship between the gut and the mind.’
The main tools of complexity science – a case study from diabetes research.
1. Group model building
2. Causal loop diagrams
3. System dynamics
4. Continuous differential equations
5. Network models
6. Agent-based models
In order to understand the dynamics underlying the rise in Diabetes type 2, Sloot and his colleagues used a wide range of tools from complexity science, which can also come in handy to untangle the mechanisms of mental disorders in an urban context.
Sloot and his colleagues started with group model building. They invited patients, care takers, public health scientists and many other experts on diabetes and asked them two questions. One: what do you think are the determinants that drive the relative rise in diabetes? And two: whom else should I address this question to? Then they invited the suggested new groups around the table and asked the same two questions. After about five rounds of interviews the system converged, and no new experts were suggested. The scientists processed the input from the interviews in correlation and causation matrices, checked it against literature and captured it in causal loop diagrams. They subsequently pruned the diagrams using sensitivity analyses that allowed them to identify the most important driving factors of diabetes development. By using longitudinal data from cohorts and assessing the temporal effects from cross sectional data they represented these factors and their underlying relationships in a system dynamic model. They subsequently created continuous differential equations that allowed them to calculate how changes in one factor might influence others in this system dynamic model. For example: how would changes in one’s socioeconomic status influence one’s physiology and eventually the risk for diabetes?
Furthermore, they used network models and differential equations to describe the dynamics of the physiological aspects and the social economical system and the interaction between these so-called ‘under and over the skin’ determinants. What do the interconnections between fast food, oxidative stress, disturbances in the hormonal balance and liver damage look like, for example? And how is this related to patient’s social economic status and access to healthy living? Subsequently, the team assigned individual characteristics to the nodes in the social environment, transforming the network model into an agent-based model. In the end the team was able to reason both qualitatively and quantitatively about the connections between lifestyle, obesity and the development of diabetes type 2.
Researchers at the CUMH perform a similar deep dive in the dynamics underlying mental health and disorder in an urban setting. ‘However,’ Sloot warns: ‘this is even more complex than the search for the drivers of diabetes, as with diabetes we at least have an exact diagnosis. We “merely” want to understand how it comes about. With mental health and mental disorders, the phenomena themselves are not reducible to a certain physiological metric, so the search for insights is even more complex, as we cannot view any aspect of it in isolation. In this holistic approach, where everything seems to be connected to everything, it is even more important to use the tools of complexity science to identify the most meaningful relationships and zoom in on these for save and reliable intervention strategies.’