Systems science methods for complex public health challenges

Systems science presents a paradigm shift; an acknowledgement that public health challenges emerge within complex systems. A move towards systems thinking has already resulted in a deeper understanding of obesity, substance abuse and smoking, contributing to whole-system intervention design.13 14 The WHO has called for systems thinking to now support further efforts to prevent non-communicable diseases.15 Yet, the uptake of systems methods in public health research has not kept pace with this development and methods often remain abstract concepts in this research field.16

The building blocks of systems are elements: individuals, households, organisations or countries. Elements are interconnected units, but may interact differently over space and time, resulting in a variety of possible outcomes at the micro level, meso level and macro level.17 As illustrated in table 1, complex systems can be defined by their properties of emergence, feedback and adaptation, and may exhibit adaptive, self-preserving, dynamic or even evolutionary traits.18 Importantly, a complex system is one whose properties cannot be fully explained by the behaviour of its component parts in isolation.19 A system can also be part of overarching systems; for example, the healthcare system can be understood as part of a wider system of public services.20 These properties of complex systems found in public health often present challenges for traditional epidemiological study designs, which tend to decontextualise the interactions of agents and overlook the potential for feedback and adaptation that occur in real-world settings.21

Cost-of-living crisis as a complex systems problem

The cost-of-living crisis presents a timely case study of a systems problem with public health implications. As a major feature, it affects disposable incomes, a key determinant of health, while limiting the opportunities for efficient population-level interventions.

The systems aspect of the cost-of-living crisis becomes apparent when trying to understand its causes. The International Monetary Fund attributes today’s high rates of inflation to multiple causes. With prices in the economy reflecting the interaction of supply and demand, the current imbalance leading to inflation is mostly thought to be due to constrained supply of goods and services.22 These patterns have emerged as part of the interaction and unintended lags of previous policy responses to COVID-19, and its effects have clearly been exacerbated by Russia’s invasion of Ukraine in 2022, and the resulting disruptions in food production and energy supplies.22 A country’s ability to adapt and respond to these shocks will also be mediated by their previous broader policy choices, for example, from where they source their energy and goods. In the UK, for example, real income was already falling behind comparator countries, falling 0.2% between 2007 and 2021 compared with an OECD average increase of 0.8%.23

It is often assumed that increasing wages in line with inflation is a straightforward solution. However, disconnecting inflation from wages is often a conscious policy decision to avoid a wage-price spiral, a vicious cycle which might fuel inflation further via increased.22 Central Banks try to temper inflation by increasing interest rates, mostly acting to dampen demand rather than the underlying supply problem directly, risking recession and increased unemployment. A simple point of intervention is therefore not straightforward when the wider system is considered, calling for a systems science approach when considering impacts of interventions on population health (see figure 1).24

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Systems science methods and cost-of-living crisis

Systems science methods can contribute to informing responses to the cost-of-living crisis. We discuss how four selected key methods (SSMs, microsimulation, ABMs and SDMs) could: (1) help conceptualise the problem and its implications for population health in a policy-relevant way and (2) help produce evidence of simulated system effects and compare trade-offs between potential policy responses from a public health perspective. Previous papers have already dealt in detail with the general introductions to these methods.6 16 A summary of each selected method is presented in table 2. An overview of each method’s contribution to public health is given in table 3.

Soft systems methodology and systems mapping

SSM is a qualitative method for a holistic framing of systems.25 It is an action-oriented process of inquiry and is frequently conducted through workshops and interviews. SSM offers a clearly defined cycle to guide users through a process of both learning, and identifying actions to be taken to improve complex, unclear (social) situations. This cycle involves identifying the problematic situation; initial drawing of Rich Pictures to illustrate the situation and the relevant elements and relationships present within it; translation of these pictures to conceptual models representing different worldviews of the variety of stakeholders involved, and possible pathways for addressing the problem; comparison of the conceptual models with the real world to gauge discrepancy; feasible and acceptable actions for improvement considered and enacted to address the problematic situation.26

Setting the system boundaries of the problem to examine can be a challenge for each of the following systems science methods discussed. Which elements of the system to focus on can quickly become unwieldy. This is particularly true for a problem like the cost-of-living crisis, which cuts across many domains and potentially affects most, or all, population subgroups in different ways. SSM can help reign in the problem, to prioritise which aspects are important for different stakeholders. For example, we might want to focus on the impact decreasing household incomes and government spending might have on child health outcomes. SSM could help identify and bring together relevant stakeholders to understand priority risks and possible (or: feasible/acceptable) interventions to address these risks. For example, this could result in a more focused and defined causal loop diagram, mapping out what problems and solutions each policy sector is working on and how things could be connected to attempt to alleviate the crisis and address specific child health outcomes. These causal loops could also subsequently be combined with other methods for quantitative simulation, outlined below.

Microsimulation

The second complexity method relevant to our case study is microsimulation. Microsimulation models can provide a virtual simplification of the real world. They use disaggregated data, often individual or household-level data, and are therefore well suited to studying impacts on health inequalities. Informed by the integration of existing theory and evidence, they combine real-world data with mathematical functions to create simulated projections for highly varied (heterogeneous) individual units. The processes simulated are a set of transition probabilities for movements between states, for example, from a healthy to a disease state.27 The models can be classified as either static or dynamic. Static models provide arithmetic predictions of morning-after-effects from a change in the system, such as a tax or benefit policy.28 29 Dynamic models, instead, provide a synthetic world to perform virtual experiments on an evolving population.27

One of the immediate concerns for the public health community might be the expected health impacts of the cost-of-living crisis, to prepare and equip future community and health system resources and prioritise future research. Microsimulation is an established method that can help inform this and allow comparison under different policy scenarios. For example, mental health outcomes might be a particular concern in the shorter term with imminent financial stress and poverty. The Heath Equity and its Economic Determinants model combines output from EUROMOD—a static microsimulation model used by the European Commission to inform decisions on the distributional impact of changes in tax and benefit policies28—with a dynamic microsimulation model to project the distributional effects of these policies on mental (and general) health outcomes over multiple years.30 A simplified version of this approach has already been used to illustrate the implications of the cost-of-living crisis, showing outcomes under different energy price cap policies for the UK.31 For a better understanding of the cost-of-living crisis, a dynamic model capturing the change in value of income in different periods could be used to model the reduction in household spending on healthcare or health benefiting activities. This could provide long-run projections for possible health effects of the crisis for public health preparation and budgeting.

Agent-based models

ABMs are computational models where agents, the unit of analysis in the system, interact with each other and their environment. These dynamic and adaptive interactions separate them from microsimulations, adding to the complexity they can model.32 The rules of engagement and behaviour of agents can be determined from simple heuristics or defined based on theory or data. Environments situate agents in a representation of space, which can either reflect actual physical space (eg, geographical information systems modelling of real cities33), or completely simulated space. As agents interact, they adapt their behaviour which has system-level consequences.34

ABMs are particularly strong for understanding how knock-on, or unforeseen, system consequences can result from changes at the individual level. They provide a possible and plausible explanation for an emergent macro-level outcome, showing how this can emerge, given a set of assumptions and constraints. For example, devalued income is likely to have impacts on spending and purchasing behaviour of individuals, which might differ substantially by population subgroup. ABMs could be used to explore how people might subsequently change their spending on unhealthy commodities (eg, alcohol, tobacco, gambling) and how these spending decisions may differ between groups depending on social influences (eg, norms) and what is available in their local environment (eg, off-licences and betting shops). ABMs could also allow examination of the mechanisms of changes in spending impacting on employment in different sectors, such as retail, and the resulting population health consequences. For example, cuts and restrictions of spending on various services (eg, cleaners, gardeners, dog walkers) and potential impacts on the affected individuals’ health through increased stress, precarity of employment and further reduced income. Examining how system effects might change with different policy scenarios is again possible by varying the environment the individuals might face, or the combination of individual circumstances that might be experienced under different public health policy interventions.

System dynamics models

SDMs enable modelling systems at an aggregate level, making them an ideal choice for testing policies or wider implications of shocks.5 SDMs are built on the idea that stock changes over time can be modelled by their interdependence on in- and out-flows, feedback loops and lag effects. One key feature of SDMs is the strong stakeholder engagement nature which lends itself to early collaboration. Like SSM, the model building process itself can aid shared understanding of the complex problem and involve relevant intersectoral stakeholders in the policy process. Like microsimulations and ABMs, SDMs allow ex ante forecasting of a systems’ behaviour and estimation of alternative outcomes given different potential public health policy choices. The simulated long time horizons are a key strength of SDMs, particularly since public health deals with upstream causes of the causes, where it is extremely difficult to directly capture lagged health effects causally with empirical research.35

Since the cost-of-living crisis is ultimately a whole-economy problem, many of the population health solutions might necessarily be explored at this aggregate level. For example, rising energy prices have been identified as a key driver of recent acute escalations in prices, particularly in Europe and the UK. Energy prices are a key input for basically every other sector, including food production and transport, and supply of any other product or service that affects population health. Unfortunately, energy also represents an input where demand, a major focus of Central Banks, is difficult to reduce without causing many knock-on consequences across sectors, for instance, unemployment with decreased production and sales. Modelling the aggregate interactions and these key drivers of current inflation might help to explore ultimately more effective public health solutions in the long term. For example, it could allow modelling of windfall taxes and other direct energy policy options, and comparison of the relative benefits of up/downstream policy efforts on stock of population health as a key outcome of interest. SDMs could also model the impacts on specific subsystems, or impacts on other social determinants of health, such as the housing market, the provision of third sector services, or health and social care capacity.

Overall, this paper has covered four of the main systems science methods and discussed how they can be applied to the cost-of-living crisis. Table 4 provides an overview of strengths and limitations of the selected methods. Importantly, these methods can be combined in hybrid-models to attempt to further dampen these limitations.