Forecasting type-specific seasonal influenza after 26 weeks in the United States using influenza activities in other countries

Abstract

To identify countries that have seasonal patterns similar to the time series of influenza surveillance data in the United States and other countries, and to forecast the 2018–2019 seasonal influenza outbreak in the U.S., we collected the surveillance data of 164 countries using the FluNet database, search queries from Google Trends, and temperature from 2010 to 2018. Data for influenza-like illness (ILI) in the U.S. were collected from the Fluview database. We identified the time lag between two time-series which were weekly surveillances for ILI, total influenza (Total INF), influenza A (INF A), and influenza B (INF B) viruses between two countries using cross-correlation analysis. In order to forecast ILI, Total INF, INF A, and INF B of next season (after 26 weeks) in the U.S., we developed prediction models using linear regression, auto regressive integrated moving average, and an artificial neural network (ANN). As a result of cross-correlation analysis between the countries located in northern and southern hemisphere, the seasonal influenza patterns in Australia and Chile showed a high correlation with those of the U.S. 22 weeks and 28 weeks earlier, respectively. The R2 score of ANN models for ILI for validation set in 2015–2019 was 0.758 despite how hard it is to forecast 26 weeks ahead. Our prediction models forecast that the ILI for the U.S. in 2018–2019 may be later and less severe than those in 2017–2018, judging from the influenza activity for Australia and Chile in 2018. It allows to estimate peak timing, peak intensity, and type-specific influenza activities for next season at 40th week. The correlation between seasonal influenza patterns in the U.S., Australia, and Chile could be used to forecast the next seasonal influenza pattern, which can help to determine influenza vaccine strategy approximately six months ahead in the U.S.

Forecasting seasonal influenza-like illness in South Korea after 2 and 30 weeks using Google Trends and influenza data from Argentina

“…In this study, total INF in Argentina showed a high correlation with ILI after 30 weeks in South Korea. In our previous study, total INF in Australia and Chile showed a high correlation with ILI after 22 and 28 weeks, respectively, in the United States [ 7 ]. Countries in the southern hemisphere showed different influenza patterns, and we identified those with seasonal patterns and influenza outbreaks similar to but preceding those of the reference country [ 7 ].…”

“…Missing data were replaced with a zero. Total influenza (INF) was defined as the sum of the number of influenza A and B viruses detected among processed specimens [ 7 ].…”

“…Cross-correlation allows for the time lag between two time-series to be identified [ 22 ]. If the higher cross-correlation value was found to have a negative lag, the values of the first series (ILI in South Korea) were correlated with the values of the second series (other variables), and the second series was made to precede the first in lag weeks [ 7 ]. The ILI in South Korea was compared to the input variables, and we selected variables with a time lag of -20 weeks or less and a correlation coefficient of 0.6 or more for seasonal forecast targets.…”

“…However, there are no such official influenza forecasting challenges and systems for forecasting seasonal targets in South Korea. Previously, we forecasted seasonal influenza patterns after 26 weeks in the United States using influenza activity in Australia and Chile, where the seasonal patterns and influenza outbreaks were similar to but preceded those observed in the United States [ 7 ].…”

Infodemiological study of COVID-19 in Latin America and The Caribbean

“…It has been used in studies about infectious diseases, mental health, general diseases, public behavior and its seasonality. [7][8][9][10][11] . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.…”