Modelling epidemics in a complex world

EMPA researcher Elise Racine works on project to identify how infectious disease modelling can best inform policymakers.

Elise Racine was working in a hospital in Dharamsala in Northern India a decade ago when high rates of tuberculosis threatened the local community. There she observed a problem for the first time that is plaguing policymakers in managing today’s COVID-19 pandemic. How to treat individuals while also treating a large-scale health problem. Who should get medical help first? How do you get people to change their social behavior on a broad scale to stem the spread of disease? How can you expect the disease to develop, depending on the answers to these questions?

During her time in India, Racine saw officials attempt to determine the implications of various policy responses aimed at reducing the tuberculosis burden. This was an eye-opening moment: It was possible to use information about the behavior of individuals and project how this might impact epidemic spread.

“My background is in sociology, but I was always passionate about public healthcare,” says the second-year Executive MPA student at the Hertie School in Berlin, who is focusing on healthcare, digitalization and big data. “A big part of why I wanted to study at Hertie was to learn how to take scientific and programmatic data and translate it into policy and communicate with policymakers.”

Since last June, Racine has been working on a project with Hertie School Professor of Ethics and Technology Joanna Bryson, Professor of Data Science Slava Jankin, Senior Research Scientist William Lowe and other researchers, which aims to find the most effective methods for modelling epidemics. To determine this, the researchers are conducting interviews and surveys with leaders in the field of infectious disease modelling and studying novel ways of using models to help improve policy.

After obtaining her BA at Stanford University, Racine worked as a counselor in California in a residential treatment center for children suffering from trauma, where she became versed in dealing with the many stakeholders involved in health problems. “That taught me how to talk to people: parents, students, police officers, case workers, healthcare professionals, teachers,” she said. “It was hard and challenging but I learned so much about the process – how many individuals and groups are involved with treating even one single individual. Then of course for a public health intervention you have to multiply that tenfold.”

One of the key issues for modellers who are trying to multiply that tenfold, or in the case of COVID-19, millionfold, is that it is hard to predict the behavior of a group comprised of many heterogeneous constituents. Many traditional methods of modelling are limited by looking at populations as homogeneous groups. One particular type of modelling the Hertie School researchers find promising for dealing with this issue is “agent-based modelling”, according to Racine. This type of modelling allows policymakers to see how decisions at the individual level – so-called agents – may impact, for example, epidemic spread.

Racine gives the example of mask-wearers. “Not everyone wears a mask,” she says. “Of those that do, some wear surgical masks and others cloth masks. All these people go out and interact with other individuals. Of these interactions, only a certain percentage follow social distancing guidelines. You need a way of looking at this complexity and representing the real world. That’s where agent-based modelling comes in. In a simulation, each person makes these decisions based on their own predefined rules, which may be influenced by their age, gender, profession, socioeconomic status etc. This can then tell you how these various actions and interactions contribute to the spread of the virus.” Agent-based modelling has been used extensively in other contexts, such as biology, ecology and the social sciences, Racine notes. “But it has real potential in epidemiology.”

The project, called “Understanding Infectious Disease Modelling”, looks at a variety of modelling methods, of which agent-based is only one. The goal is to make this knowledge and methods more accessible for policymakers, to discuss different options and be more critical of their limitations, Racine says.

“We have found in talking to people that they struggle to find resources and information, and that limits their knowledge,” she says. To improve knowledge-sharing, the research group has set up a website,, where people involved in making policy can find out about technical and design choices that go into building models and the associated advantages, disadvantages, and policy implications.

The researchers hope to uncover trends in how modelling methods are being used, offer an overview of infectious disease modelling, identify resources, highlight best practices and understand how modelling is incorporated into policymaking.

For agent-based models a whole variety of expertise may be tapped – experts in data collection, cognitive intelligence, psychologists and other fields, including policymakers, Racine says. “This is something that the Hertie School really provides – asking policymakers what knowledge they need to make decisions. Models are forecasts. They offer an idea of what might happen in the future. Models can disagree and reality may not match them. Policymakers need to understand why there is this disconnect and what insights can be still be drawn.” 

Racine, who also holds an MSc from the London School of Economics and Political Science (LSE) in Health and International Development, will incorporate some of the research into her Hertie School EMPA master’s thesis, and plans to begin her doctorate in Population Health at Oxford University this fall.

“I was so excited to have this opportunity at Hertie and it’s been fascinating,” she says. “This project helped me identify one of the areas I am passionate about – which is how tools are used for health policy. No matter what avenue I go down, this project has made me realise I want to tackle similarly challenging problems, and I am really grateful for that opportunity.”

The Hertie School is not responsible for any content linked or referred to from these pages.
Views expressed by the author/interviewee may not necessarily reflect the views and values of the Hertie School.