A model to predict SARS‐CoV‐2 infection based on the first three‐month surveillance data in Brazil

Diaz‐Quijano, Fredi A.; da Silva, José M.N.; Ganem, Fabiana; Oliveira, Silvano; Vesga‐Varela, Andrea L.; Croda, Julio

OBJECTIVE: COVID‐19 diagnosis is a critical problem, mainly due to the lack or delay in the test results. We aimed to obtain a model to predict SARS‐CoV‐2 infection in suspected patients reported to the Brazilian surveillance system. METHODS: We analyzed suspected patients reported to the National Surveillance System that corresponded to the following case definition: patients with respiratory symptoms and fever, who traveled to regions with local or community transmission or who had close contact with a suspected or confirmed case. Based on variables routinely collected, we obtained a multiple model using logistic regression. The area under the receiver operating characteristic curve (AUC) and accuracy indicators were used for validation. RESULTS: We described 1468 COVID‐19 cases (confirmed by RT‐PCR) and 4271 patients with other illnesses. With a data subset including 80% of patients from Sao Paulo (SP) and Rio Janeiro (RJ), we obtained a function which reached an AUC of 95.54% (95% CI: 94.41% ‐ 96.67%) for the diagnosis of COVID‐19 and accuracy of 90.1% (sensitivity 87.62% and specificity 92.02%). In a validation dataset including the other 20% of patients from SP and RJ, this model exhibited an AUC of 95.01% (92.51% – 97.5%) and accuracy of 89.47% (sensitivity 87.32% and specificity 91.36%). CONCLUSION: We obtained a model suitable for the clinical diagnosis of COVID‐19 based on routinely collected surveillance data. Applications of this tool include early identification for specific treatment and isolation, rational use of laboratory tests, and input for modeling epidemiological trends.