Analysis and forecasting of the evolution of COVID-19 death numbers in the state of Pernambuco and Ceará using regression models

• Autor(a) principal: Santos, André Luiz Pinto dos
• Data de Publicação: 2020
• Outros Autores: Figueiredo, Marcela Portela Santos de, Ferreira, Tiago Alessandro Espínola, Gomes-Silva, Frank, Moreira, Guilherme Rocha, Silva, José Eduardo, Freitas, Jucarlos Rufino de
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Research, Society and Development
• Texto Completo: https://rsdjournal.org/index.php/rsd/article/view/4551
• topic: Coronavírus; Previsão; Modelagem Epidêmica; Pandemia.; Coronavirus; Forecast; Epidemic Modeling; Pandemic.; Coronavirus; Pronóstico; Modelado epidémico; Pandemia.

The purpose was defined to adjust different non-linear models in the analysis to death data by COVID-19 in Pernambuco and Ceará and to extrapolate the deaths numbers through forecasts. In this report, we analyze the official epidemic data available by the Ministry of Health of Brazil (MS), referring to the period of 25/03/2020 to 11/05/2020 for Pernambuco - PE and in the period of 26/03/2020 to 11/05/2020 for Ceará, of the deaths numbers, COVID-19 confirmed. For the comparison between the models, the adjusted coefficient of determination (), residual mean squares (RMS), and Akaike information criterion (AIC) were used. All models had good adjustments, with values of approximately 99%. The verification of the assumptions of the residues was carried out through graphic analyzes, and the assumptions were met. The cumulative deaths’ numbers in the period from 12/05/2020 to 10/10/2020 was calculated for Pernambuco and 12/05/2020 to 11/10/2020 for Ceará, in addition to the extrapolation of the absolute growth rate (AGR) for the respective intervals. The analyzes indicated that the inflection points of all models occurred within 200 days after the start of the pandemic. However, it is not yet possible to make reliable projections of when the numbers of confirmed deaths will minimize. Regardless of the possible uncertainty of the models' prediction, our observations indicate that the next few days may be critical in determining the future growth of death cases.