Covid19Analytics
Short-term forecasting tool for COVID-19 cases and deaths in Brazil
With the objective of creating a tool to support the management of the health crisis by the responsible authorities, as well as informing the public, a group of professors from PUC-Rio built a model to predict the number of cases and deaths of COVID-19 in Brazil. up to two weeks ahead (Medeiros, Marcelo, et al., 2022). The founding professors, who are associated with different departments (Electrical Engineering, Industrial Engineering, and Economics), have quantitative articles in several areas published in top international journals, in addition to extensive experience in the construction and practical implementation of predictive models for different sectors through its laboratories.
The model, which is driven only by data, estimates a relationship between Brazil and countries that were affected by the pandemic in a previous period. The fact that Brazil is “lagging behind” in relation to these countries allows, based on this estimated relationship, to project the evolution of the number of cases and deaths in the short term. The use of this model is complementary to the use of epidemiological models, which are difficult to quantitatively discipline because it is a new disease, whose scientific community still has little knowledge.
The proposed model is re-estimated every day based on new data from the Brazilian trajectory and from other countries, which are automatically selected based on their adherence to the trajectory and predictive capacity. So far, the model has only been tested for official data reported by the Ministry of Health (https://covid.saude.gov.br), and relevant indicators of testing and social distancing have not yet been considered.
The model provides the prediction of the evolution of the total number of reported cases, number of new cases, as well as the growth rate. The same prediction method is also applied to deaths, and could be applied to specific cases of critically ill hospital networks. In addition to point predictions, based on means, the model also informs a 95% confidence interval, which takes into account the risk arising from the prediction error.