Estimating underdiagnosis of COVID-19 with nowcasting and machine learning

Abstract

ABSTRACT: Objective: To analyze the underdiagnosis of COVID-19 through nowcasting with machine learning in a Southern Brazilian capital city. Methods: Observational ecological design and data from 3916 notified cases of COVID-19 from April 14th to June 2nd, 2020 in Florianópolis, Brazil. A machine-learning algorithm was used to classify cases that had no diagnosis, producing the nowcast. To analyze the underdiagnosis, the difference between data without nowcasting and the median of the nowcasted projections for the entire period and for the six days from the date of onset of symptoms were compared. Results: The number of new cases throughout the entire period without nowcasting was 389. With nowcasting, it was 694 (95%CI 496–897). During the six-day period, the number without nowcasting was 19 and 104 (95%CI 60–142) with nowcasting. The underdiagnosis was 37.29% in the entire period and 81.73% in the six-day period. The underdiagnosis was more critical in the six days from the date of onset of symptoms to diagnosis before the data collection than in the entire period. Conclusion: The use of nowcasting with machine learning techniques can help to estimate the number of new disease cases.