Detect the spread of coronavirus through keywords on Google

Micrograph of SARS-CoV-2 virions isolated from a patient during the 2019-20 coronavirus pandemic (NIAID)

In the past three months, more than 1.3 million people worldwide had been infected with SARS-CoV-2, the virus behind the COVID-19 outbreak. Utilizing technology that had evolved, people continue to look for information about the virus.

In the United States, the keywords "what is the coronavirus?" were popular on Google. Then, there were also the keywords "o que fazer em caso de suspeita de coronavirus?" or "what to do if a coronavirus is suspected?" which was popularly typed in Brazil. In line with that, "corona vaayaras tips" alias "tips to prevent coronavirus" were sought by Indians.

Seth Stephens-Davidowitz, the author of Everybody Lies: Big Data, New Data, and What the Internet Can Tell Us about Who We Really Are, in his opinion column in The New York Times, stated that Google's search data, specifically about a disease or outbreak, could actually be converted into estimates of the spread in locations where disease-related searches were high.

From the end of March to the beginning of April, searches for "I can't smell" increased in New York, New Jersey, Louisiana, and Michigan. According to Stephens-Davidowitz, there was strong evidence that people infected with SARS-CoV-2 experienced loss of the sense of smell, AKA anosmia, whose numbers were estimated to affect 30 to 60 percent of sufferers.

In the US, the regions where the search for "I can't smell" increased had a high number of coronavirus infections. In New York, for example, there were more than 140 thousand total residents infected with more than 5,5 thousand people killed due to COVID-19.

Then, the search for "no puedo oler," which had the same meaning as "I can't smell," was also often sought by Ecuadorians. For your information, Ecuador was one of the countries with the highest coronavirus infection rates in South America.

Still referring to Stephens-Davidowitz, Joshua Gans, a professor at the Rotman School of Management, the University of Toronto, stressed that the search for "non sento odori" was sought by Italians shortly before COVID-19 devastated the country.

An Italian flag with the slogan "Andrà tutto bene" (Pietro Luca Cassarino)

Vasileios Lampos, a researcher at University College London, in his paper titled Tracking COVID-19 Using Online Search, explained that until March 24, 2020, keyword searches related to the symptoms of COVID-19 were indeed popular in the communities, especially in the US, Canada, Britain, Australia, France, Italy, and Greece.

In general, popular keywords on Google that could be used to detect the spread of coronavirus were "I can't smell," "I feel fever and chills," and "my eyes hurt." According to Lampos, the high search by keyword for COVID-19 symptoms had a prevalence that was comparable to reality in the field.

The prevalence between Google search and the reality on the ground did not only occur when the worldwide pandemic of COVID-19 outbroke. In the past, Google had a special tool called Google Flu Trends (GFT) to map the spread of flu in the community through searches related to flu symptoms.

Mauricio Santillana, in a study entitled What Can Digital Disease Detection Learn from (an External Revision to) Google Flu Trends? mentioned that GFT worked by detecting flu-related keywords. There were 45 terms that were later published by the Centers for Disease Control and Prevention (CDC).

The problem was, according to Santillana, that the GFT methodology mapping the flu in the community was inaccurate enough. For example, the 45 CDC's flu-related terms used by GFT were more of a CDC's prejudice, not the empirical fact that the 45s were closely related to flu.

In addition, the flu was one of the diseases that had many derivatives. As a result, the terms also varied and developed over time. There were swine flu, H1N1, to H1N9, which unfortunately were not used by GFT.