Using Twitter to Track Disease: Weighing the Advantages and Challenges

A few weeks ago I participated in a fantastic twitter chat on the use of social media for public health.  During the event, our moderators posed the following question: "Are there any other diseases (besides the flu) that we could track on social media?"

The question generated a very lively discussion that I was inspired to revisit on Storify this morning after reading the Washington Post's article, "Twitter becomes a tool for tracking flu epidemics and other public health issues." 

The WP article highlights several advantages and challenges of monitoring public health diseases and/or conditions on twitter.  My twitter chat colleagues brought up many other important issues for us to consider, so I'm including these expanded lists:

Advantages:

• Offers real time data on health or behavior (Government data can often take weeks or months to be released)
• There is so much available data! 
• It could capture cases that would not otherwise be formally documented at a physician's office or hospital.
• It has proved helpful in tracking time sensitive disease outbreaks (e.g., Novovirus).  *Check out this article about how twitter was used to track Norovirus activity during a journalism conference.

Challenges:

• Accuracy and case definition (i.e., does a twitter user really have the flu or just a bad cold?)
• Tracking specific words like "sick" or "flu" can bring up a lot of content that is unrelated to the twitter user being ill themselves (e.g., "I'm so sick of this terrible weather").  *Check out how Johns Hopkins researchers are working to address this problem by better screening tweets.
• We must differentiate between tracking symptoms vs. tracking cases- they are not the same.
• Our search strategies should include various terms or slang that are used to describe the disease or behavior of interest.
• Caution: media coverage of certain illnesses can cause a spike in key words on twitter without a rise in actual cases.
• What are the privacy concerns?
• Twitter might not thoroughly capture diseases or conditions that carry stigma (e.g., mental illness) because users may be hesitant to discuss them in a public forum.
• Results could be skewed by populations who are over or under represented on twitter.
• Do we need to train "trackers" to intervene? E.g., what if they are monitoring dangerous tweets/behaviors like suicidal ideation and attempts?

While the challenges list is quite long, I hope we are not discouraged!  I think twitter is an enormous resource for public health professionals.  We just need to be thoughtful and thorough regarding how to use twitter effectively.

More Resources:

The Washington Post article and related stories shared great links to more information about research in this area:

• YouTube video on University of Rochester efforts to track influenza on twitter.  It also describes their related app: Germ Tracker (warning: it may have you hopping off your regular morning bus). 
• Johns Hopkins University article: "You Are What You Tweet: Analyzing Twitter for Public Health".
• Brigham Young University article: "'Right Time, Right Place' Health Communication on Twitter: Value and Accuracy of Location Information".
• A great article that highlights what we can learn from Google Flu (since their predictions were off this year)- emphasizes the importance of "re-calibrating" your models or algorithms each year.

What Do You Think?

• What other advantages and/or challenges should we add to the list?
• What other resources can you share?

  

More about → Using Twitter to Track Disease: Weighing the Advantages and Challenges

Google Continues to Use its Power for Public Health Good

Yesterday, Google announced its new surveillance system for Dengue Fever. Dengue Fever is a disease caused by four related viruses spread by a particular species of mosquito. It can cause high fever, rash, muscle and joint pain, and in severe cases- bleeding, a sudden drop in blood pressure (shock) and death. Millions of cases of Dengue infection occur worldwide each year. Most often, dengue fever occurs in urban areas of tropical and subtropical regions.

The system is similar to that which was previously released as their Google Flu Trends program. These systems use search queries within Google (for example those that enter the disease's name and/or symptoms) to identify trends. The Dengue system also takes advantage of a new feature called Google Correlate, which shows previously unknown correlations between search terms. These correlations allow researchers to model real world behaviors by examining internet search trends. For those who may be skeptical of this model, you should check out a publication (co-authored by Google and the Centers for Disease Control and Prevention-CDC) in the 2009 Nature Journal . The article reports that "because relative frequencies of certain queries were highly correlated with the percentage of physician visits in which a patient presents with influenza-like symptoms, we can accurately estimate the current level of weekly influenza activity in each region of the United States, with a reporting lag of about one day."

This is a pretty exciting addition to public health surveillance (where the goal is systematic, ongoing, data collection that is used to monitor trends, identify priorities, direct resources, identify emerging hazards, and evaluate interventions).

This is not the first time that Google has jumped into the public health field with an impressive contribution. In 2010, Google searches related to suicide started appearing with a message guiding users to the toll-free number for the National Suicide Prevention Lifeline. The number is 1-800-273-8255. Triggered by searches such as "I want to die" or "ways to commit suicide," the number is listed next to an icon of a red telephone, at the top of the search results.

The addition of the Lifeline number came shortly after (at the suggestion of a Google user), the company started displaying the hotline for the American Association of Poison Control Centers after searches for "poison emergency."

These cases of Google's work in public health are great examples of effective health communication and public health principles:

• Identifying the primary channels through which your audience searches for health information (more and more are utilizing the internet) and delivering accurate and effective information and/or interventions via those channels.

• Maximizing data driven surveillance systems- using existing data (e.g., internet searches) to identify public health trends.

• Building strong partnerships (as evidenced by the publication by Google and CDC) CDC has partnered with a company with specific expertise and resources in an area that can be invaluable to their work.
  

More about → Google Continues to Use its Power for Public Health Good