Background: Real-time surveillance of search behavior on the internet has achieved accessibility in measuring disease activity. In this study, we systematically assessed the associations between internet search trends of gastrointestinal (GI) symptom terms and daily newly confirmed COVID-19 cases at both global and regional levels.
Methods: Relative search volumes (RSVs) of GI symptom terms were derived from internet search engines. Time-series analyses with autoregressive integrated moving average models were conducted to fit and forecast the RSV trends of each GI symptom term before and after the COVID-19 outbreak. Generalized additive models were used to quantify the effects of RSVs of GI symptom terms on the incidence of COVID-19. In addition, dose-response analyses were applied to estimate the shape of the associations.
Results: The RSVs of GI symptom terms could be characterized by seasonal variation and a high correlation with symptoms of "fever" and "cough" at both global and regional levels; in particular, "diarrhea" and "loss of taste" were abnormally increased during the outbreak period of COVID-19, with elevated point changes of 1.31 and 8 times, respectively. In addition, these symptom terms could effectively predict a COVID-19 outbreak in advance, underlying the lag correlation at 12 and 5 days, respectively, and with mutual independence. The dose-response curves showed a consistent increase in daily COVID-19 risk with increasing search volumes of "diarrhea" and "loss of taste".
Conclusion: This is the first and largest epidemiologic study that comprehensively revealed the advanced prediction of COVID-19 outbreaks at both global and regional levels via GI symptom indicators.
Keywords: COVID-19; Gastrointestinal symptoms; Google trends; Time-series analysis.
Copyright © 2021. Published by Elsevier Ltd.