A modified SEIR model to predict the COVID-19 outbreak in Spain: simulating control scenarios and multi-scale epidemics

Leonardo R Lopez; Xavier Rodo

The projected rise in the number of infections shown in Figure 1 is very sensitive to the latest data used, as these data are the ones governing the changes in trends, therefore errors in reporting or other large uncertainties associated with these values may dramatically alter the final outcome in terms of number of infected and associated mortality. For the curve to happen as shown, the simulation assumes the initial conditions hold, therefore if that were not the case and measures were lifted before the epidemic reaches R 0 ≈ 1, infections might rise again. Improvements can be achieved with a more complete integration of time delay coordinates in classical SEIR models [6] . This way, both the incubation period and the period before recovery, as well as the precise role of asymptomatic population can be better represented at times when these information is highly uncertain. A substantial gain in the epidemic projection in the form of reduced infections in the population would already be achieved with a four-times rise in the efficacy of the control situation we had around March 9 − 16. This was the time when social distancing/confinement was imposed (March 13). Time will show to which extent current measures manage to increase the value of alpha. Application of strict containment measures clearly show drastic effect on the epidemic progression and a substantial reduction in both infected and to a lesser extent, deaths ( Figure 5 ).