| Abstract [eng] |
Relevance and novelty of the work: Traditional epidemiological indicators are published with delays, limiting timely responses to emerging outbreaks. International studies suggest that Google Trends may provide earlier signals of infectious disease spread, but evidence from Lithuania—especially in the post pandemic period—is limited. The aim of the study is to assess the possibilities of using Google Trends in predicting the incidence of COVID-19, influenza and other acute upper respiratory tract infections (AURTI). The tasks of the study are to assess the possibilities of using Google Trends in predicting the incidence of (1) COVID-19, (2) influenza, (3) AURTI in Lithuania. Methods: An ecological time series study analysed weekly COVID 19, influenza, and AURTI incidence in Lithuania alongside Google Trends search intensity for “cough”, “flu”, “coronavirus”, and “covid”. Graphical inspection and cross correlation analysis were applied to identify temporal associations. For keywords showing significant correlations, ARIMA models integrating historical incidence and Google Trends indicators were developed to evaluate their contribution to short term forecasting. Results: For COVID 19, a significant correlation was identified only for the keyword “coronavirus”, with search intensity increasing 2–4 weeks before recorded morbidity (strongest correlation at a four week lag: r = –0.165, p = 0.017). Other keywords showed no significant associations (p > 0.05). An ARIMA(2,1,0)(0,0,1)[52] model demonstrated that including Google Trends data reduced COVID 19 forecasting errors. For influenza, significant correlation was found only for the keyword “flu” (lags –5 to +2 weeks), with the strongest association at lag 0 (r = –0.661, p < 0.05). An ARIMA(1,1,1)(0,1,0)[52] model showed improved forecasting accuracy when Google Trends data were included. For AURTI, significant correlations were observed for “cough” (lags –6 to +3 weeks; strongest at lag –6: r = –0.184, p = 0.001) and “flu” (lags 0, 1, 2, 5; strongest at lag 0: r = –0.237, p < 0.05). However, ARIMA models (ARIMA(1,0,1) for “cough” and ARIMA(1,0,3)(2,0,0)[52] for “flu”) showed that incorporating Google Trends increased forecasting errors. Conclusions: 1. Google Trends search intensity indicator for the keyword “coronavirus” may help in refining the forecast of COVID-19 incidence. 2. Google Trends search intensity indicator for the keyword “flu” may help in refining the forecast of influenza incidence. 3. Although Google Trends search intensity indicator for the keywords “cough” and “flu” is related to the incidence of OVID, its usefulness for forecasting is questionable. Keywords: Google Trends, incidence, forecasting, COVID-19, flu, AURTI. |
