Skip to main content Skip to navigation

Modelling the impact of a pandemic and the resoponse to it on socially transmitted health beaviours

Alongside direct deaths from a pandemic, wider health impacts may be considerable and could even exceed the direct effects of the infection. Non-communicable diseases (NCDs) are responsible for 71% of all deaths globally and nearly 90% in the UK as well as substantial chronic morbidity. Four key modifiable behaviours increase risk: tobacco use, physical inactivity, unhealthy diet and alcohol abuse. Effective targeting of these behaviours may prevent multiple NCDs simultaneously. There is evidence that these behaviour ‘spread’ through social networks. A pandemic and responses to it such as lockdowns that disrupt social networks may provide risk and opportunity for substantial behavioural change.

An agent-based model (ABM) is a microsimulation in which agents (representing individuals, organisations, etc.) act according to pre-specified decision rules. An agent’s attributes, state, and environmental conditions act as inputs to the decision rules, which determine the agent’s behaviour, including its interactions with other agents, and may update its attributes and state. An ABM takes individual data and parameters and allows the emergence of population-level phenomena that differ from a simple aggregate of individual behaviours. Thus, ABMs can be used to discover complex causal effects, identify underlying mechanisms behind complex systems, and help to make sense of large amounts of existing evidence and data. Recent reviews of ABMs in public health and in NCDs specifically have found that ABMs are under-utilised as a tool to give insight into the epidemiology of NCDs. The few existing ABM studies of NCDs are predominantly focused on a single health behaviour, with the remainder focused on a single disease. Our aim is to capture the essence of individual-level decision making in terms of a set of rules and environmental factors, model the links between the risk factors and development of NCDs, and evaluate the impact of disruptive events such as pandemics and the associated response on these. Project supervisors (Dr Oyebode WMS and Dr Griffiths Computer Science) have created and validated a synthetic network generator with supporting community detection algorithms, resulting in a set of ~250K agents and their connections (e.g. representing an urban population equivalent to a small city or London borough) based on identified or estimated parameters from available population data.

This project would aim to

  1. Develop an ABM in which agents express NCD risk behaviours and develop disease informed by real-world epidemiological evidence. Both environmental parameters, and factors internal to the individual, such as genetics and knowledge, will contribute to an individual agent’s likelihood of developing an NCD. We will consider an agent’s internal model, and how this may change through interactions with its environment and with other agents.
  2. Perturb the model to artificially change the environment simulating ‘lockdown’ or other interventions which restructure social contacts for the purposes of pandemic response, to examine effects on disease development and population level prevalence, and identify theoretical targets for public health intervention.
  3. Validate and disseminate the model’s output by involving public health experts, both policy-makers and practitioners, throughout the project.

The novelty of this work comes from the focus on using real-world incomplete data to create an ABM with a focus on validating that this model provides realistic results comparable to the populations it is based on. Our model will consider the contribution of multiple NCD risk-factors which will interact and evolve over time, even without intervention; and it will provide a higher level of autonomy to the individual agents than existing epidemic models.