Google Trends provides access to an unbiased sample of Google searches. The Google Trends interface reports a “query share,” calculated by dividing the number of queries of interest by the total number of queries for all search terms over the same time period and region. Each query share is normalized on a scale of 0 to 100, with 100 representing the maximum value of the share for the period and region selected [14]. The scaled query share values are plotted daily, generating a time series.

The chosen search terms were gastrointestinal symptoms that have previously been reported to be associated with COVID-19 infection in the literature, including diarrhea, nausea, vomiting, and abdominal pain. The terms fever and cough were included as positive controls. The term constipation was included as a negative control, as we felt this symptom was unlikely to be associated with COVID-19. The terms anosmia, dysgeusia, loss of appetite, loss of taste, and loss of smell were considered; however, due to the low frequency of searches for these terms, analysis was limited by missing data. The default “All categories” and “Web search” settings were selected for the Google Trends query.

Daily case counts of confirmed COVID-19 cases for each US state were obtained from the Johns Hopkins University Center for Systems Science and Engineering data repository [15].

Daily query shares for the selected symptoms were obtained from October 1, 2019 to June 15, 2020 for the United States. The full data set of search query shares is provided in Multimedia Appendix 1. The data were divided into two time periods for comparison: a baseline period during which the COVID-19 case burden was low (October 1 to February 29) and a post–COVID-19 period (March 1 to June 15). The query share for each symptom was divided by its average for the pre–COVID-19 period to generate a curve of search interest relative to the pre–COVID-19 baseline. To examine longer-term patterns, the search query shares for the 5-year period preceding the COVID-19 pandemic were plotted. A 3-day moving average smoother was applied to reduce day-to-day variation. Cumulative and new COVID-19 cases from the United States were superimposed on Google search data to assess their temporal relationship with the symptoms. All analyses were performed with Stata 13.0 (StataCorp LP).

Our analysis of aggregate internet search query data reveals that the search query shares for symptoms associated with COVID-19 rose in advance of the substantial increase in identified cases that occurred with the first wave of the pandemic in the United States in early March 2020. The data suggest that symptoms of fever, cough, and diarrhea may occur contemporaneously and precede case identifications by up to two weeks in the United States, particularly during the early weeks of the pandemic. This study validates the findings of Higgins et al [16] that COVID-19–related internet searches preceded case identification by over a week in China, Italy, Spain, Washington, and New York.

This study also suggests that there was no significant increase in abdominal pain or constipation queries, which may provide reassurance to clinicians who are faced with these very common complaints in the setting of a new and uncertain pandemic.

The seasonal increase in search query shares for fever, cough, and diarrhea in December 2019 and at the same time in prior years can be attributed to increased search interest during the typical cold and influenza season in the winter. These query shares are much lower than those seen during the post–COVID-19 time range in this study.

Despite the consistent increase in cumulative case count throughout April and May, our findings show that search queries for fever, cough, and diarrhea begin decreasing in mid-March, when the new daily case rate was over 5000 and continuing to rise. There are several possible explanations for the decoupling of COVID-19 cases and search query interest. One explanation is that users sought information via the internet early in the pandemic when there was less public knowledge regarding the virus and its manifestations and that by April, the demand for further information was saturated. During the early weeks of the pandemic, access to outpatient medical care and COVID-19 testing were limited; however, later in the pandemic, both testing and access to telehealth visits became more common, and individuals may thus have relied on alternative sources of information. Our study suggests that internet search query data can provide early clues to the start of an outbreak but may have less utility as the course of the pandemic extends.

There are many limitations and assumptions that must temper our interpretation of these data. Through this infodemiological approach, data were only gathered from internet users, who may not reflect the entire population, such as younger or older persons. Moreover, individuals may be searching for these terms for reasons other than being symptomatic themselves. The role of media attention in influencing user behavior should also be considered. However, public knowledge of the gastrointestinal symptoms associated with COVID-19 was minimal during the period in which the search volumes rose and peaked, which suggests that search interest in diarrhea was less likely to be influenced by media reporting of diarrhea as a manifestation of the disease.

This study demonstrates sharp increases in internet search interest in fever, cough, and diarrhea at the onset of the COVID-19 pandemic in the United States preceding case identification. Further work is warranted to determine if infodemiological approaches can contribute to population-based surveillance of early outbreaks.