The next phase of the global battle against COVID-19 is in full swing, and huge testing facilities to monitor the rate of transmission are up and running. An entirely new simulation system can very accurately assess the development of epidemics and the effects of potential interventions. However, if the models are to work people cannot choose whether they want to be tested. This should be a civic duty so that we can get an accurate picture, the researchers say, and every year – just like in the military – major exercises should be held to assess whether we are prepared for future pandemics.
Around the world, ordinary people during the COVID-19 lockdown have been transformed into amateur statisticians trying to understand and follow the development of the pandemic. Both the general public and decision-makers have very much needed some type of control and security. Where are we on the infection curve? Is the rate rising or falling? Using modelling and simulation, researchers have developed a system for controlling transmission and optimizing possible interventions against COVID-19.
“The goal was to limit the prevalence of COVID-19 infection so that the epidemic would be as brief as possible and avoid exceeding the capacity of the healthcare system. With the new system, we are equipped to monitor the transmission and to assess the effects of individual initiatives. However, getting the models to work requires better testing data. Today we only get data from ill people or those who choose to be tested, which distorts the numbers. The public should carry out their civic duty to be tested regularly so that we can monitor trends and intervene when necessary,” explains John Bagterp Jørgensen, Professor, Department of Applied Mathematics and Computer Science (DTU Compute), Technical University of Denmark, Kongens Lyngby.
Cement factories and artificial pancreases
John Bagterp Jørgensen has never focused on epidemics before, but he has worked on control systems for cement plants, energy systems, bioreactors and constructing an artificial pancreas and other diabetes-related technology. This background in constructing mathematical models piqued his curiosity when Denmark’s Minister for Health and Senior Citizens Magnus Heunicke presented the notorious green and red curves at a press conference showing two possible trajectories for the epidemic: one in which the capacity of the health system could manage the epidemic and one in which the capacity was exceeded.
“I thought: wait a minute. This may well become an elongated green curve, and it instantly occurred to me that this is exactly what we can do with our models and systems: calculate when the curve hits the maximum capacity and what is necessary to prevent this from happening. So I immediately started crunching the numbers, and these quickly confirmed that this COVID-19 crisis will not be over soon and that an enormous effort is needed to determine what works and what does not,” says John Bagterp Jørgensen.
The idea was born. As a mathematician and model builder, John Bagterp Jørgensen had never worked on epidemics and viruses but an artificial pancreas that can automatically dispense the right amount of insulin to people with type 1 diabetes. However, the tools were more or less the same. Using stochastic models, researchers can systematically build models and calibrate them as new data arrive.
“We use stochastic models to describe the evolution of a phenomenon over time in which random events, as known from virus epidemics, play a crucial role. With our models, we can therefore continually simulate, optimize and recalibrate, so that we can provide short-term forecasts, for example, about the expected number of hospital inpatients or the effects of physical distancing – at the population level, but also according to, for example, geography or social class,” explains John Bagterp Jørgensen.
Together with partners from the Department of Electronic Systems at Aalborg University and colleagues from DTU Compute, he set out to build a simulation system that could continually adjust and find the best strategy for restrictions in lockdowns and reopening after epidemics and thus protect the population and society and ensure that the capacity of Denmark’s hospital system is not exceeded.
At the same time, Kim Guldstrand Larsen, a professor of computer science at the other end of Denmark, also wanted to help – only with a slightly different solution to the urgent problem.
“We use agent-based models, which we have previously used to analyse various complex systems ranging from software in cars to biological systems. By analysing complex systems through machine learning, we can develop calculation models that simulate the system. Running the simulations many times refines the model, and then we can study the result with statistical tools and thus ensure a better basis for politicians and public health authorities making decisions about initiatives and easing of restrictions,” explains Kim Guldstrand Larsen, Professor, Distributed, Embedded and Intelligent Systems, Department of Computer Science, Aalborg University.
DTU Compute’s system uses mathematical modelling and optimizes several possible interventions, whereas Kim Guldstrand Larsen’s system uses the UPPAAL tool, which excels at describing a trend that suddenly transitions to new conditions: for example, schools and childcare institutions suddenly reopening. The researchers rapidly realized that they should join forces, and the synergy provided by the two different methods has so far been a blessing for the project.
“We have really benefited from collaborating. We have been able to continually compare and adjust the models and have also been able to work closely with the Danish Health Authority and Statens Serum Institut,” says Kim Guldstrand Larsen.
Civic duty to be tested
The two new modelling systems now provide an important instrument so that when COVID-19 flares up again and in the other epidemics of the future, the public authorities can better predict the effects of closing or opening schools, colleges and universities, isolating institutions such as nursing homes with residents who are especially vulnerable to COVID-19 and can give recommendations to vulnerable people for self-isolation when transmission is especially high in a specific area. However, a crucial factor required for these systems to work is testing and the data from these tests.
“Optimizing our models requires data that are representative of the entire population. The problem with the test data arriving right now is that they either come from people who are ill or from people who choose to be tested because they want to be certain that they are not infected. What we need are data from a random sample of the population. So if our systems are to provide the right answers, the public authorities should test widely, and being tested should be a civic duty,” explains another participant in the project, Henrik Madsen, Professor and Head of Section, DTU Compute, Kongens Lyngby.
With valid data from virus testing and antibody testing, researchers will be able to use the new model systems to identify new waves of epidemics very accurately before they gain traction so that they can be tamed.
“We probably could not have used our models to predict the recent outbreak in Hjørring, Denmark, but we could help to monitor whether it would dissipate and whether new outbreaks are on the way, as long as we get the necessary data,” says Henrik Madsen.
Another prerequisite for the new modelling systems to function is training and optimization.
“To ensure that the modelling systems are up to date and that we know how to use them, we should conduct annual exercises, like the military, simulating an epidemic so that we know when and how to respond,” he explains.
The researchers also think that the simulation system can be used to prepare the population for future epidemics – through games.
“One of our plans is to develop this into a computer game so that the public can experience some of the mechanisms at play when an epidemic develops. Experiencing which factors are crucial for transmission may also affect personal behaviour,” says Henrik Madsen.
The Novo Nordisk Foundation has awarded a grant of DKK 4,969,057 for the project Estimation, Simulation and Regulation for Optimal Interventions Related to COVID-19 in a collaboration between the Technical University of Denmark and Aalborg University. The Poul Due Jensen Foundation awarded Kim Guldstrand Larsen and Jakob Stoustrup a grant of DKK 1,205,233 for the project BEO-COVID.