Adaptive SIR-models’ Ability to Both Estimate the Reproduction Number and Forecast the Future Spread of Covid-19 in Sweden

• Main Authors: Fröberg, Tobias, Nordström, Ludwig
• Format: Others
• Language: English
• Published: KTH, Skolan för elektroteknik och datavetenskap (EECS) 2021
• Subjects: Computer Sciences; Datavetenskap (datalogi)
• Online Access: http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-302493

In outbreaks like the Covid-19 pandemic, mathematical models can be used to estimate various epidemiological parameters which model the evolution of the disease. A popular model that is used by Folkhälsomyndigheten (FHM) for estimating the effective reproduction number (Rt) has the shortcoming that it cannot simultaneously forecast the future number of cases. This thesis explores an extension of another model, the SIR-model, in which the model parameters are fitted to recorded data. This makes the model adaptive, opening up the possibilities for estimating the Rt daily and making predictions of future number of confirmed cases. The thesis use this adaptive SIR-model (aSIR) to estimate the Rt and create forecasts of new cases in Sweden. The thesis’ purpose is to determine how precise aSIR-models are at estimating the Rt (when compared with FHM’s model). It will also analyze how accurate aSIR-models are at simultaneously forecasting the future spread of Covid-19 in Sweden. The results of this thesis showed an aSIR-model is good at estimating the Rt, with a similar precision when compared to the Rt estimated by FHM’s method. The aSIR-model also quickly adapted to sudden changes and dynamics during the outbreak in a similar way to that of the method used by FHM. When used as a tool for forecasting, the aSIR-model performed well for shorter time periods or when the Rt had a low variance. For longer time periods or when the Rt varied a lot, the forecast accuracy fell significantly.