Internet searching may predict novel symptoms, but not new outbreaks. A long-term evaluation of infoveillance in COVID-19

Background: At the beginning of COVID-19 pandemic authors in several countries reported the possibility of predicting disease outbreaks using internet analysis and search tools like GoogleTrends™. Our aim was to investigate the impact of changes in COVID-19 symptomatology and pandemic intensity on those predictions. Material and methods: GoogleTrends™ was utilized to track online searches for COVID-19 symptoms in Poland during two years of the pandemic. Search volumes were then assessed for correlation with daily cases in each wave of infection separately. Results: The symptoms that correlated strongly with new cases were anosmia and ageusia (Spearman's rho=0.5230 and rho=0.4483 respectively, p<0.01). Searches for these symptoms preceded an outbreak by 12 days during the first wave of infections, but this gap was later shortened to five days. The frequency of searching for these symptoms markedly diminished during the last phase and was no longer adequate. Stronger correlations were then shown for fever, sore throat, and headache. Conclusions: In conclusion, COVID-19 case prediction using GoogleTrends™ did not remain possible later on in the pandemic course. However, noticeable changes reflecting novel features of emerging SARS-CoV-2 variants were observed. Therefore, monitoring symptom changes and virus evolution might be a promising application of internet search analysis in the future.