So, what is the use of epidemiological models?

BLOG: Heidelberg Laureate Forum

Laureates of mathematics and computer science meet the next generation
Heidelberg Laureate Forum

What, then, is the use of epidemiological models? That was a central thread running through Tuesday’s hot topic panel on “Mathematics of Disease – the Science of Epidemic Modelling.”

And the panel was an interesting mix of different perspectives on that question, including modelers, namely Sebastian Funk of the London School of Hygiene and Tropical Medicine and Sheetal Silal of the Modelling and Simulation Hub at the University of Cape Town, but also practitioners like Amrish Baidjoe from Doctors without Borders, and Julia Fitzner, who leads the data management and acquisition team for the COVID19 response in the WHO Health Emergencies Programme.

All panelists agreed, of course, that epidemiological models are not predictions of the future – although, lamentably, they have been misrepresented as such in at least some of the media coverage of the Covid-19 crisis. Just how problematic that can be came as something of a shock at the onset of the pandemic, or, as Sebastian Funk put it: this was “communicating the hard way”, namely by having the public misunderstand what was presented often enough before scientists learned to attach suitable content warnings, “notes of caution” to their publications.

Several examples made very clear some of the fundamental limits of models: Changes in human behaviour – which may be in reaction to news coverage, or to a general sense of urgency or non-urgency drawn from personal assessments of the current level of severity of the pandemic – are virtually impossible to include in epidemiological modelling in a suitable way, and neither are policy decisions that can change the conditions for the spreading of a disease in drastic ways.

So what, then,  is the use of models? There, the answers of the panelists added up to a fairly coherent picture, but with some interesting nuances. Sheetal Silal pointed to the fight against malaria, where the South African government had used models to find out, among other things, the cost of their policies in different scenarios, and to identify possible funding gaps.

Julia Fitzner pointed to models used by the WHO to try and estimate whether closing borders and different kinds of lock-downs, were of help in the fight against the Corona pandemic.

Amrish Baidjoe characterized models as informing us about different scenarios, telling us about the different ways matters might possibly turn out. But when asked about concrete examples from his own work, he admitted that for him, in deciding about the proper measures “on the ground”, epidemiological studies were usually more helpful than models. Sometimes, he said, models could be a distraction – in particular when the scientists creating the models were not sufficiently in touch with the people on the ground.

In the end, it probably makes sense to look at epidemiological models as the somewhat more sophisticated cousins of simple statements such as “if infections were to continue rising exponentially at the current relative rate”: ways of quantifying how the epidemic will continue under certain explicit assumptions (in the simplest case, “if conditions remain as they are now”, but also with assumptions like “what if we manage to get 50%, 70%, 90% of our people vaccinated”). A projection, the answer to a what-if, where information about the what-if itself should always be included, lest people take what they are seeing for a prediction.

This is probably the most important information about epidemiological models that scientists should spread. And the HLF Panel “Mathematics of Disease – The Science of Epidemic Modeling” drove that point home quite forcefully.

Markus Pössel hatte bereits während des Physikstudiums an der Universität Hamburg gemerkt: Die Herausforderung, physikalische Themen so aufzuarbeiten und darzustellen, dass sie auch für Nichtphysiker verständlich werden, war für ihn mindestens ebenso interessant wie die eigentliche Forschungsarbeit. Nach seiner Promotion am Max-Planck-Institut für Gravitationsphysik (Albert-Einstein-Institut) in Potsdam blieb er dem Institut als "Outreach scientist" erhalten, war während des Einsteinjahres 2005 an verschiedenen Ausstellungsprojekten beteiligt und schuf das Webportal Einstein Online. Ende 2007 wechselte er für ein Jahr zum World Science Festival in New York. Seit Anfang 2009 ist er wissenschaftlicher Mitarbeiter am Max-Planck-Institut für Astronomie in Heidelberg, wo er das Haus der Astronomie leitet, ein Zentrum für astronomische Öffentlichkeits- und Bildungsarbeit. Pössel bloggt, ist Autor/Koautor mehrerer Bücher, und schreibt regelmäßig für die Zeitschrift Sterne und Weltraum.


  1. Yes, epidemiological models can tell you how it affects (quote) 50%, 70%, 90% of our employees to be vaccinated
    or, for example, how it affects the number of malaria cases to reduce the mosquito density in a certain area.

    These types of epidemiological studies are closely related to possible interventions in disease in humans. Each type of disease intervention has its costs and difficulties and today there is often more than one method of intervention and the question arises as to which are more effective and which are less costly for a given rate of disease reduction. For example, there are new malaria vaccines that reduce the likelihood of infection and the question arises: what is more effective and cheaper? 1) Is it reducing mosquito density through the use of insecticides or genetically modified mosquitoes or 2) is it a vaccination that offers partial protection against the disease or 3) would a combination of 1) and 2) be the best?

Leave a Reply

E-Mail-Benachrichtigung bei weiteren Kommentaren.
-- Auch möglich: Abo ohne Kommentar. +