Predicting COVID-19 in very large countries: The case of Brazil

• Published in: PloS one Vol. 16; no. 7; p. e0253146
• Main Authors: Parro, V C, Lafetá, M L M, Pait, F, Ipólito, F B, Toporcov, T N
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 01-07-2021; Public Library of Science (PLoS)
• Subjects: Algorithms; Biology and Life Sciences; Brazil - epidemiology; Computer and Information Sciences; Coronaviruses; COVID-19; COVID-19 - epidemiology; Delivery of Health Care; Disease susceptibility; Disease transmission; Dynamic models; Electrical engineering; Epidemic models; Epidemiology; Evaluation; Figure of merit; Humans; Machine Learning; Medicine and Health Sciences; Models, Statistical; Optimization techniques; Pandemics; Parameters; People and places; Physical Sciences; Population; Public health; Quality assessment; Research and Analysis Methods; SARS-CoV-2 - isolation & purification; Severe acute respiratory syndrome coronavirus 2; Software; System dynamics; Time series; Brazil; China
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0253146

This work presents a practical proposal for estimating health system utilization for COVID-19 cases. The novel methodology developed is based on the dynamic model known as Susceptible, Infected, Removed and Dead (SIRD). The model was modified to focus on the healthcare system dynamics, rather than modeling all cases of the disease. It was tuned using data available for each Brazilian state and updated with daily figures. A figure of merit that assesses the quality of the model fit to the data was defined and used to optimize the free parameters. The parameters of an epidemiological model for the whole of Brazil, comprising a linear combination of the models for each state, were estimated considering the data available for the 26 Brazilian states. The model was validated, and strong adherence was demonstrated in most cases.