NYU Researchers Use Google Search Data to Predict COVID-19 Outbreaks

With the currently increasing rate of Covid-19 cases, the new waves of the infection have not followed previously observed patterns of spread.Google

However, researchers from Courant Institute of Mathematical Sciences, a section of New York University, have found a way of predicting the behavior of infections long before they start surging. Interestingly, the researchers found out that this could be done by analyzing online searches.

Read Also: The WHO Accuses Rich Countries of Undermining COVID-19 Vaccines Supplies to Poor Countries

The researchers discovered that whenever there was a surge in searches for activities that could increase people’s chances of contracting Covid-19, there was always a surge in Covid-19 cases, usually after 10-14 days. Also, they noticed a reduction in infections when people searched for activities they could do while staying at home.

Anasse Bari, one of the professors who carried out this study, noted that a similar method is being used in finance, specifically to estimate businesses’ income.

Megan Coffee, a professor at New York University’s Division of Infectious Disease & Immunology, also noted that the methods used for the study could be applied to stop the spread of any infections. She went further by saying the techniques would help predict where disease outbreaks are most likely to occur.

This will also help policymakers and epidemiologists have accurate information about what usually causes infection surges, and how they can manage the situation.

Read Also: The WHO Accuses Rich Countries of Undermining COVID-19 Vaccines Supplies to Poor Countries

The main idea of the study is to create a system that uses anonymized data, such that the issue of privacy is never a problem.

Mobility Index and Isolation Index

In a bid to generate a meaningful result, the researchers had to divide keywords and search phrases into two different categories. The idea was to create a trackable system.

This categorization produced the isolation index and the mobility index.

Keywords such as “flight tickets,” “movie theaters near me”, and other keywords about activities that people could do outside their homes fell into the mobility index.

On the other hand, keywords that suggested activities that people could do at home fell into the isolation index.

The researchers also used certain keywords sourced from a popular public opinion survey, Democracy Fund + UCLA Nationscape. The public opinion survey was conducted to know what respondents would love to be doing if the government yielded to the advice of healthcare professionals by lifting bans and restrictions.

Read Also: EU Ready to Finance a Production Plant in Morocco to Meet Vaccine Demand in Africa

Among other things shown by the survey, the researchers found out that people missed going to the movies, attending a concert, and going to a sporting event.

Anasse Bari noted that the categorization was an important step they needed to take in order to effectively predict behaviors that could increase Covid-19 cases.

The Net Movement Index

The researchers were able to develop the isolation index and the mobility index using google trends to find trends in the data collected. The collection of search data happened in all the states in the United States from March to June 2020.

Since it was pertinent to show the relationship between the two indexes, the researchers developed what is known as the “Net Movement Index.”  An increase in the Net Movements shows that more mobility search queries are being made. Conversely, a decrease in the Net movement would mean that people are doing more searches on things to do at home which could likely be used to predict a drop in infections 2 weeks later.

Read Also: Do Adenovirus-Based COVID-19 Vaccines Increase the Risk of HIV Infection?

Helping Policymakers

This study also allowed the authors to further research how the removal of stay-at-home orders could affect the infection rates and the mobility index. However, the researchers only used five US states: New York, California, Arizona, Florida, and Texas for this study.

They discovered that there was a reduction in infection rates during the lockdown period and an increase in infection rates when the stay-at-home order was relaxed.

According to one of the authors of the study; the goal of the study is to show that machine learning can help humans predict behavior in future disease outbreaks.

Read Also: Moderna Is Working on a New Formulation of Its Vaccine to Combat the South African Variant of COVID-19

References

COVID-19 early-alert signals using human behavior alternative data

FEEDBACK:

Conversation

Want to Stay Informed?

Join the Gilmore Health News Newsletter!

Want to live your best life?

Get the Gilmore Health Weekly newsletter for health tips, wellness updates and more.

By clicking "Subscribe," I agree to the Gilmore Health and . I also agree to receive emails from Gilmore Health and I understand that I may opt out of Gilmore Health subscriptions at any time.