Tracking COVID-19 using taste and smell loss Google searches is not a reliable strategy

Abstract

Web search tools are widely used by the general public to obtain health-related information, and analysis of search data is often suggested for public health monitoring. We analyzed popularity of searches related to smell loss and taste loss, recently listed as symptoms of COVID-19. Searches on sight loss and hearing loss, which are not considered as COVID-19 symptoms, were used as control. Google Trends results per region in Italy or state in the US were compared to COVID-19 incidence in the corresponding geographical areas. The COVID-19 incidence did not correlate with searches for non-symptoms, but in some weeks had high correlation with taste and smell loss searches, which also correlated with each other. Correlation of the sensory symptoms with new COVID-19 cases for each country as a whole was high at some time points, but decreased (Italy) or dramatically fluctuated over time (US). Smell loss searches correlated with the incidence of media reports in the US. Our results show that popularity of symptom searches is not reliable for pandemic monitoring. Awareness of this limitation is important during the COVID-19 pandemic, which continues to spread and to exhibit new clinical manifestations, and for potential future health threats.

Figure 1

Data from which correlation was derived. Here shown for Italy (11–17 March) and the US (1–7 April), the weeks with highest correlation for taste loss and smell loss. The graphs show the taste loss (in orange) and smell loss (in green) search queries. Each point represents a different region/state. Normalized number of new cases related to the corresponding week on the x axis, popularity of the search terms for the same week on the y axis. For both Italy and the US graphs, not all the regions or states are shown because of the lack of popularity index for some regions/states.

Figure 2

Correlation over geographical regions. For the weeks of 11–17 of March, 1–7 of April, 13–19 of May and 19–25 of August (in this order), correlations between the number of new COVID-19 cases (normalized with respect to 1,000,000 inhabitants) and popularity index of each search term are shown as round points. Confidence intervals (90%) calculated for each week separately are shown as error bars.

Figure 3

Sliding windows correlation data (time frame of 31 days). Correlation values between the total number of new COVID-19 cases in Italy and US, normalized with respect to 1,000,000 inhabitants, and search queries popularity index. Conservative confidence intervals of 90% are shown as ribbons. Correlation values including the dates in which a second increase in the number of new cases in the US is shown in a purple rectangle.

Figure 4

Media coverage. Comparison between Google searches volume for taste loss and smell loss queries, Media Cloud popularity of the same keywords and the number of new COVID-19 cases. Calculation was performed from March 4th to August 25th for Italy and the US. The Media Cloud data were normalized as for the Google search results, assigning the value of 100 to the day with the highest popularity, with the other days assigned values relative to that day. The number of new cases is relative to a population of 1,000,000 in the respective country. Purple rectangle highlights 11–17 March week in Italy, and 1–7 April week in the US, the arrows indicate the first peak for smell/taste loss media popularity.