European experts are working closely together with their American counterparts on calculating future scenarios for Covid-19. One of them is Elizabeth Lee, an epidemiologist specialised in infectious diseases at the Johns Hopkins Bloomberg School of Public Health. She has been participating in the US ‘COVID-19 Forecast Hub’ as well as the US-focused ‘COVID-19 Scenario Modeling Hub’. Although the pandemic has highlighted the value of mathematical modelling for public health, she underlines the need to improve how modelling results are communicated.
How are mathematical models useful for controlling infectious disease?
Models allow us to explore possible scenarios and the impact of potential interventions so we get a sense of how an infectious disease could spread. Building a mathematical model that mimics some of the features of a dynamical system, such as disease transmission in which there are feedback loops and many interacting processes, can help us understand situations that we can’t easily explore in an experimental way.
How has your work changed since Covid-19?
I am an infectious disease epidemiologist and my primary work is in cholera. Before Covid-19 started, our team was working on mapping the burden of cholera globally and assessing the impact of cholera vaccination campaigns. We used models to try and understand how best to allocate cholera vaccines and to target them to the highest-risk populations to achieve the greatest health impact.
After Covid-19 was announced, I shifted to working full time on developing a mathematical model to understand the dynamics of how the virus might spread. It was a general framework that we were able to apply to different locations, not just within the US, but also in international settings.
Over the course of the pandemic our work has evolved from thinking about how to prepare for hospital surges, towards forecasting and trying to understand the impact of potential scenarios in the future.
How accurate are the models?
It is not like forecasting the weather – this is purely shaped by physical systems that we understand. With infectious diseases we are getting into the business of how people might behave or change their behaviour in response to a measure or policy, which is not something we can confidently project into the future.
For short-term forecasting we can make some pretty good assumptions and get a decent understanding of what is going to happen in the next few weeks. But when we try to look further into the future, I would not say it is a forecast anymore but rather a ‘scenario’ that makes assumptions about relatively realistic possibilities, so decision makers can plan and compare the potential impact of specific interventions.
How are they being used?
I think they are extremely useful for situational awareness, providing an idea of what is happening right now and what is about to happen. Like checking the traffic or the weather in the morning, it will not necessarily dictate what you are going to do, but it gets you thinking whether you need to take an umbrella or a different route to work.
It’s hard to say if a specific forecast or scenario has led directly to a policy change. We work with different agencies – presenting our mid-to-long term modelling outputs helps them to think through some of the problems and assess what measures might be more effective at dampening virus transmission.
How have disease models shaped communications about the pandemic?
I doubt people understood what R0 was before Covid-19! At the very beginning, people were talking a lot about ‘flattening the curve’, but I don’t think it was being interpreted in the right way. Taking action to flatten the curve doesn’t mean that the disease is going away, you are just spreading out when people are getting ill so healthcare systems aren’t overwhelmed. I think this led to a backlash as people thought that restrictions would be lifted once the curve was flattened.
A lot of people in our field have really stepped up on social media, made appearances in the news and written opinion pieces to improve understanding of some of these terms and concepts. Nevertheless, there are still challenges in the communication of risk and modelling results to the public.
There is a tendency in the media to pick up models with nice dashboards because they are easy to access, even when they may not clearly communicate their assumptions, limitations, and intended uses. Models are designed to answer specific questions, but it’s easy to take model projections out of those contexts. As a field we need to think more about having some standard ways of talking about mathematical models, trusted sources and pre-prepared explanations for communications to the public.
What will happen to forecasting initiatives post-Covid-19?
Joint forecasting efforts are not new, they have been used for influenza and Ebola. However, I think Covid-19 has really changed the public mindset as people have seen the value of models. Bringing modellers together and having centralised sources of data for modelling, offers new collaboration opportunities as well as the chance to promote clearer communication.
It will be really interesting to look back and see how different types of models have performed over time. We may find that models with certain features performed better at certain points in the pandemic. We will be able to tweak them to perform better for future outbreaks of Covid-19 or other infectious diseases.