Modelling the COVID-19 epidemics in Brasil: Parametric identification and public health measures influence

Renato Machado Cotta; Carolina Palma Naveira-Cotta; pierre magal

The copyright holder for this preprint (which was not peer-reviewed) is . https://doi.org/10.1101/2020.03.31.20049130 doi: medRxiv preprint Though the three parameters estimation provides a fairly good reproduction of the behaviour of the CR(t) curve for Brasil, one may observe a change in the pattern of the evolution around day 30, that could not be entirely followed by the proposed model. It is also a known fact that the initial amount of kits for blood testing that were purchased by the Brazilian government were finished around this time, and before being fully supplemented, there could have been a reduction on the number of executed exams of the symptomatic individuals, that might have affected the partition of reported to unreported cases by the end of this period covered by the present dataset. Therefore, the more general model including the time variation of the partition f(t), eqs.(4.c,d), is here implemented for a more refined inverse problem analysis. It is then expected that a reduction on the f value can be identified ( < 0 ), with an abrupt variation on the exponential behaviour, here assumed as a sharp functional time dependence (large ). Therefore, an additional statistical inverse problem analysis is undertaken, this time for estimating five parameters, namely, 0 , , 0 , , and , aimed at improving the overall agreement with the CR(t) data behaviour, with a possible reduction on the partition of the reported and unreported the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.