Experimental feature

Listen to this article

Experimental feature

Graham Dodge has reason to be grateful for catching a stomach bug. It was while he was laid low with the virus that he came up with the idea for Sickweather, a website that uses social media to track the spread of disease.

“I went on Facebook to see who among my friends had the same thing,” says the US computer designer and entrepreneur. “That’s when I realised that this basic need to connect and see who else was sick could be done on a much larger scale, using Twitter and Facebook.”

Sickweather, which Mr Dodge launched with James Sajor and Michael Belt in 2011, tracks the spread of diseases, producing something akin to a weather map for common ailments. It is part of a growing trend by scientists and health organisations to make use of the mass of information generated by social media.

New York’s University of Rochester, for example, runs a project to monitor flu using Twitter, the microblogging site. And in a sign that perhaps political leaders are starting to take notice, the US Department of Homeland Security, has engaged Accenture, the management consultancy, to help it investigate how social media network data could be used to better inform and protect the public in the event of an infectious disease outbreak or a biological attack.

Mr Dodge’s creation of Sickweather followed his experience launching an online crime map. He says Twitter is well suited to tracking diseases. Sickweather covers 27 health problems including cold, flu, stomach virus, chickenpox, eczema and insomnia.

Academics, too, confirm that illness-related tweets from Twitter’s more than 500m users worldwide can be revealing. Researchers at the University of Iowa found that estimates of flu-like illness derived from Twitter accurately tracked reported disease levels during the pandemic of 2009. A study by the University of Bristol reported similar results in the UK.

In a paper entitled You Are What You Tweet: Analyzing Twitter for Public Health, Michael Paul and Mark Dredze from Johns Hopkins University, Baltimore, concluded that Twitter provides “quantitative correlations with public health data and qualitative evaluations of model output. Our results suggest that Twitter has broad applicability for public health research”.

They added: “Twitter can have a greater impact on public health informatics than just influenza tracking . . . millions of users could provide new tools for public health research.”

Internet users’ searches also offer a snapshot of disease trends. Google says results from its Flu Trends initiative closely match the data collected by the US Centers for Disease Control and Prevention. The search engine group runs a similar project to track dengue fever activity in Latin America and south Asia.

Even though social media are a rich source of intelligence on the spread of disease, they are not yet a substitute for more conventional methods of gathering information. One issue, say experts, is the provenance of the data.

“Evaluating data for epidemiology is not just about taking raw data,” says Ken Eames, a lecturer at the London School of Hygiene and Tropical Medicine. “The problem with social media information is that it is often anonymous, and you may not know so much about its source, so the value for epidemiology may be limited.”

Mr Eames is involved in running Flusurvey.org, a website that tracks the spread of the illness throughout the UK. Rather than monitoring social media, Flusurvey relies on registered users to supply details about when and where they are sick. The project, now in its fourth year, has sister sites that cover the Netherlands, Belgium, Portugal, Italy, Austria, Germany, Switzerland, France and Sweden, so data from all the countries can be compared.

Mr Eames says: “The advantage of [an online system] is that it does not dependent on anyone going to hospital to report their infection, which is the way this sort of information has traditionally been collected. And we get a lot of very useful extra information, such as whether people went to see a doctor, whether they took time off work and whether they took medicines.”

Nevertheless, Mr Eames is an advocate of using social media to gather health data. “The great amount of data available from Twitter is bound to be a great resource for tracking diseases,” he says. “Another value of something like Twitter is that it can get just about everywhere to gather data, and get an idea of people’s location.”

At Sickweather, Mr Dodge believes there is plenty more that can be done with social media data to learn about disease. “We really want to open up to other languages, as right now we are only using English,” he says.

He also believes social media could be harnessed to provide valuable insights into illnesses such as cancer.

“We would like to see what social media can reveal that has not been revealed yet from other studies,” Mr Dodge says. “How environment impacts on chronic illness, for example, or other factors such as weather and travel. I think a lot could be learnt from comparing other data to the information we’ve gathered from social media.”

Copyright The Financial Times Limited 2019. All rights reserved.

Follow the topics in this article