The numbers from the Covid-Sim computer model are said to have prompted British Prime Minister Boris Johnson to make a sharp political U-turn in March.

The model of Neil Ferguson and colleagues at Imperial College London predicted 500. 000 Corona deaths in Great Britain if the government did not take appropriate countermeasures.

Such models as the Covid computer model are as useful .Sim are also used to assess the effectiveness of containment measures, however, they have a similar weakness as climate models: Since the calculations are highly non-linear, the results can vary widely depending on the parameters used and the input data. Peter Convey and colleagues therefore analyzed the sensitivity of the model on behalf of the London Royal Society. The result, which they published on the preprint platform Researchsquare at the beginning of November, shows how sensitive Covid-Sim reacts to small changes in its inputs.

How sensitive is the model? The Coveney team found 500 parameters in the Covid-Sim code, of which 19 had the greatest influence on the result. Up to two-thirds of the differences in the model’s results were due to changes in three key variables: the length of the latency period during which an infected person has no symptoms and the virus cannot pass on; the effectiveness of social distancing; and how long a person goes into isolation after infection.

In order to better quantify the sensitivity of the model, the researchers recommended that epidemiologists use statistical methods that are also used in the assessment of climate models: Small changes in the starting conditions and parameters are systematically carried out during simulations and the totality of the results is statistically evaluated. In order to make the computational effort manageable, the models are mostly simplified for this purpose – which Convey and fellow campaigners also recommend for the Covid simulations.

Static basic assumptions Similar to physical climate models, the spread of contagious diseases can be described mathematically with a set of coupled differential equations. The individual equations indicate the change over time of people who are susceptible (S) to the virus, have become infected (I) and then either recover (R) or die – and do not spread the virus any further. Such models are therefore also called SIR models.

The simplest SIR models are based on quite static basic assumptions, such as that everyone has the same chance of contracting the virus from an infected person infected because the population is perfectly and evenly mixed. More realistic models divide people into smaller groups – according to age, gender, state of health, occupation, number of contacts, etc. However, many of the parameters can only be estimated. For example, the modelers from Imperial College assumed for their study around March that 0.9% of people infected with COVID – 19 would die and that people who show no symptoms can spread the virus for 4.6 days after infection.

(wst)