– It feels absolutely fantastic, when I got the phone call I thought: is this even possible, considering that last year’s awardees were also from Mathematical Sciences, says Torbjörn Lundh. Even if they only followed the criteria for the award, I think it was brave of the committee.
– Yes, and the motivation was of course very enjoyable to read for us, Philip Gerlee adds (see fact box).
It is now almost exactly 18 months since the World Health Organization stated that the corona virus outbreak had become a pandemic. A few days before that, the logisticians Ingrid Fritzell and Julia Karlsson, who work with the chief physician Thomas Brezicka at Sahlgrenska University Hospital, had contacted Philip Gerlee and asked for a meeting that same day. The time factor was important, the usual academic procedure would not be fast enough.
“What do you want?”
– We felt that we wanted to contribute for real and be a cog in fighting the pandemic, and asked: what do you want? And it turned out to be models for predicting how many people would be admitted to hospital, and when those patients would appear. We had prior knowledge of related subjects as modelling and biological data that came to good use, and first we began to build simple models, then increasingly refined that were adapted to Sahlgrenska University Hospital. Then they also wanted us to evaluate other models, for example there was an early Icelandic model. (Read more: They predict the need for care for covid-19 patients, interview from September 2020).
After a month Torbjörn and Philip wrote a debate article: All models are wrong – some are useful (in Swedish, Alla modeller är fel – några är användbara, Svenska Dagbladet April 22, 2020). There were many strong opinions about how Sweden handled the pandemic and they wanted to explain why one must be careful with complex models. The article was noticed by epidemiologists in Linköping and Lund, who wanted to collaborate. Clinicians, epidemiologists and modellers compared their research results, which among other things resulted in a project on how travel patters can predict hospital admissions.
The spread of diseases depends, among other things, on the number of physical encounters between people, and mobile phone data as well as data from public transport utilisation were used as indicators for this. Both turned out to capture the pandemic waves well, and this can also be used to predict the need for medical care (Read more: Travel patterns can predict care needs during the pandemic, from January 2021). Philip was also a co-author of a report commissioned by the Public Health Agency of Sweden, which evaluated the pandemic models of PHAS and of others during 2020. (Read more: Uncertain pandemic models still helped understanding, interview from May 2021).
Many possible indicators
During a second and third wave of infection, completely different models are needed than in the beginning. Among other things, Torbjörn and Philip are now trying to estimate the infectivity of novel virus variants. Basically it is evolutionary biology, which also is knowledge that they have. They are also trying to find more indicators that can contribute to the models. There are many possibilities, but unfortunately they are often stopped by legal obstacles. A pretty good indicator turned out to be calls to 1177, linked to certain diagnostic codes as cough. Mobility data was also useful, though it was much easier to obtain data from Västtrafik than from the mobile phone companies. Analyses of wastewater could perhaps give an even earlier signal.
– We have both previously worked together with clinicians and cell biologists, but epidemiologists were new to us and they had entirely different requirements when it came to what it means for a model to be useful and how model valuation is carried out. It has been very educational to get acquainted with the healthcare culture, for example working with such sensitive data that you only may use certain laptops for it. It is both useful and exciting to move across culture barriers.
In two weeks, Torbjörn and Philip will participate in a Nordic conference arranged by Nordforsk and the research network NordMathCovid. Many of the participating researchers are linked to their countries’ public health authorities and make predictions for them, and here possible future collaborations can be formed. They have also applied for money within the Swedish Research Council’s call for emergency preparedness research, the reply is expected in December. The plan is to continue to evaluate all the different pandemic models that exist and see how to connect them with decision makers in the best possible way. With better use of surveillance models for spread of infections, there may be a toolbox ready to use before the next pandemic.
Text and photo
: Setta Aspström