What can Google search trends for COVID-19 symptoms tell us about the direction of the pandemic?


While studies have indicated that Google search data can be a useful tool, an expert says it’s best to consider this data alongside other more reliable metrics before drawing conclusions about the direction of the pandemic. .

Jagpreet Chhatwal, an assistant professor at Harvard Medical School and associate director of the Institute for Technology Assessment at Massachusetts General Hospital, who created a simulator for COVID trends, said multiple metrics can be considered together to show that levels of COVID-19 are decreasing in the State. The MGH model predicts that Massachusetts is now peaking at the Omicron surge, while other models have said the state has peaked, or will soon. The level of COVID-19 in Boston-area wastewater is also falling, and Google search data shows “a very similar trend.”

“Look at Google Analytics, the prediction models and the sewage all pointing to the same trends,” Chhatwal said. “That tells us something about what’s happening in the state, which is that the trends should be down from now on.”

A study published in the scientific journal Nature in July 2021 examined whether internet searches for COVID symptoms can predict the spread by analyzing Google trends in Turkey, Italy, Spain, France and the UK from January 2020 to August 2020.

This study found that internet search interest in COVID symptoms “is a reliable predictor of later reported cases for the first wave of the COVID-19 pandemic”, and that search interest via Google can be “used to alert the healthcare system to prepare and allocate needed resources in advance. The study identified that there was a lag between testing for COVID symptoms and new cases that varies by country.

The study also noted that experts disagree on the validity of using Google Trends data as an epidemiological tool, but their research concluded that when studying the relationship between data on real-world trends and data like COVID-19 cases, “the resulting correlation may be more reliable”.

In September 2020, Google made search trend data available for researchers to “investigate the link between symptom-related searches and the spread of COVID-19” in the hope that the data “could lead to a better understanding of the impact of the pandemic”. .”

Google Trends data includes trends for over 400 symptoms and health conditions across US counties and also contains trends for the past three years to account for search changes due to seasonality. The data represent the volume of searches per symptom compared to a benchmark of what is typical for a given region.

Chhatwal said Google Trends data is best used in tandem with other metrics because it depends on human behavior.

For example, if a new COVID-19 symptom were to appear, searches might increase for that particular symptom. Once the symptom becomes familiar among the population, Google searches will decrease even though cases may continue to rise, Chhatwal said.

“It’s not always that Google Trends is going to track exactly what’s going on with actual epidemiology,” Chhatwal said.

In another example, if a family member is infected with COVID, a person may be looking for a particular symptom at that time. Later, if they or another family member are infected, they can no longer search because they have already searched for it, Chhatwal said.

“Research is not as highly correlated as other things,” Chhatwal said. “Once you have the knowledge, what’s the point of doing more Google searches? That’s what I wonder. This is where things can get a little jarring.

Sewage, on the other hand, is directly related to the amount of virus in the community, so if COVID levels are rising or falling, “we should expect a direct correlation,” Chhatwal said.

The most recently detected levels of COVID-19 in the Greater Boston area continued their downward trend, falling to less than a quarter of their Omicron-fueled peaks in early January. Biobot Analytics, which performs the tests, said it found the amount of virus detected correlated with newly diagnosed coronavirus cases several days later.

“Ballpark, starting this week, we should start to see things calm down starting next week. That’s what most models project,” Chhatwal said.


Previous Globe material has been used in this report.


Amanda Kaufman can be contacted at [email protected] Follow her on Twitter @amandakauf1.

Previous Software as a service improves daily IT operations
Next Windows 11 Outlook search not finding recent emails? Try this workaround