Researchers found association between prevalence of COVID and search queries, posts
University of Waterloo
Using the example of COVID-19, researchers found there
was an association between the disease’s prevalence and search engine queries
and social media posts.
“The general public tends to use internet searches and social media for health information, and especially so during global epidemics,” said Dr. Yang (Rena) Yang, a postdoctoral research fellow in the School of Public Health Sciences at Waterloo.
“These behaviour patterns can be used by public health authorities to develop a real-time surveillance system to flag when diseases are spiking or waning or respond quickly to emerging infectious diseases.”
The team extracted symptom keywords from Google Trends and Twitter data in Canada from January to March 2020. These keywords included cough, runny nose, sore throat, shortness of breath, fever, headache, body ache, and fatigue on Google Trends.
On Twitter, researchers looked at
COVID-19-related hashtags, such as pneumonia, cough, fever, running nose and
breath. They then cross-checked the information against COVID-19 data from the
COVID-19 Canada Open Data Working Group.
The researchers found that search terms related to
COVID-19 symptoms strongly correlated with daily COVID-19 cases with a time lag
of between one and 13 days, suggesting that these tools can serve as early
warning signals for digital disease surveillance in real time. The
sophisticated machine learning model used for forecasting in this study
performed better with Google Trends than with Twitter data.
Dr. Zahid Butt, lead investigator of the
study and an assistant professor in the School of Public Health Sciences at
Waterloo, noted there are challenges in modelling due to the noise from
self-generated data, not to mention the ability to identify relevant keywords
of an emerging infectious disease.
“Our future research will aim to systemically identify and organize pertinent symptom keywords for emerging diseases, even before they are commonly recognized or reported,” Butt said.
“These systems have the
potential to assist in epidemiological control and monitor public perceptions
of the disease, as well as forecast trends in outbreaks. A multifaceted strategy
that uses multiple data sources and multimodal modelling would help provide
accurate and comprehensive emerging disease surveillance.”
The study, Digital
Disease Surveillance for Emerging Infectious Diseases: An Early Warning System
Using the Internet and Social Media Data for COVID-19 Forecasting in Canada,
appears in Studies in Health Technology and Informatics and was authored by
Waterloo's Dr. Yang (Rena) Yang, Shu-Feng Tsao, Mohammad Basri, Dr. Helen Chen
and Dr. Zahid Butt.