Building pandemic models on the fly: how to develop data-based models when time is at a premium and good data are scarce
Location: CECAM-HQ-EPFL, Lausanne, Switzerland
In 2020 the world was faced with a global pandemic: COVID-19, which has claimed millions of lives. The response to this pandemic has involved medics, scientists and engineers. Parts of this response, for example the rapid development of vaccines, have been triumphs of modern science. Data analysis has also helped understand which policy interventions have been effective in containing the spreading of the epidemic [1-3]. For example, it was found that multiple measures are necessary to effectively contain the epidemic, with no single policy being sufficient on its own . However, on the whole, these insights came too late to influence policy.
Epidemiological models such as the Susceptible-Exposed-Infected-Recovered (SEIR) model , can predict the short-time evolution of the number of infected individuals, if one is able to estimate the parameters entering in the model. These approaches are based on a population-wide description, and can be considered a mean-field approximation of a network description, but there are many, more refined models that account for the heterogeneity of society. In network models  one can account quantitatively for the fact that each individual has a finite set of contacts to whom they can pass the infection.
One crucial feature of network models is that they can account for the overdispersion (non-Poisson distribution) of the number of secondary infections. The standard SEIR models miss this important aspect. Yet overdispersion in itself conveys information about the mode of transmission. Typically, airborne viruses exhibit overdispersion .
In all these models parameters are typically inferred from the epidemiological data. For example, a key parameter is the transmission probability of the pathogen. But transmission of a virus is in part a physical transport process, out of one body, across air and into another. This workshop will consider this transport process from a physical science modelling perspective.
Transmission of the virus SARS-CoV-2 is complex. There is by now, overwhelming evidence that airborne transmission is dominant . An infected person breathes out aerosol particles, some of which contain SARS-CoV-2, which can partially evaporate and remain suspended in the air. These virus-carrying particles can then be inhaled by a susceptible person, who may become infected. The probability that someone becomes infected depends not only on biological factors, but also on physical processes.
These physical factors are studied across disciplines such as soft-matter physics, fluid-mechanics and aerosol-science. Physics determines how micrometre-sized droplets can form in our lungs, our vocal chords, or the liquid lining of our respiratory tract. Physical processes determine how virus particles are incorporated into droplets of various sizes and compositions, how they are transported as aerosols in air and how long the nanoscale virus capsid can survive in an evaporating droplet. Transport of aerosols is well understood [9,10]. However we sorely lack direct information or models for the formation of virus-containing droplets, and for the processes that limit the lifetime of viable virus.
Despite the interdisciplinary nature of the models needed here, little interdisciplinary research is being done. This leaves us no better prepared for the next pandemic than we were in 2019. What we need, but don’t have, is transferable (from one disease to another) models, which need to be based in part on the physical processes (eg droplet formation, aerosols) that are common to airborne disease transmission.
Alessandro Laio (SISSA) - Organiser
Daan Frenkel (U Cambridge) - Organiser
Richard Sear (University of Surrey) - Organiser