Can Twitter Track Suicide Risk?
How useful is Twitter in identifying people at risk for suicide?
Posted Feb 03, 2014
Suicide is the tenth leading cause of death in the United States according to the most recent statistics. In 2009 alone, there were 36,909 suicides and those statistics likely don't account for the total number of deaths (many are classified as accidental). The statistics on suicide also don't include the enormous number of self-inflicted injuries with an estimated 472,000 emergency room visits in 2007 alone. Along with the deaths or injuries associated with suicide, there is also the enormous physical and emotional toll that suicide places on the ones left behind, whether family or friends.
Gathering suicide data is always difficult and makes research into suicide prevention far harder than it needs to be. Identifying people at risk for suicide requires being aware of suicide threats as they occur even though they are often ignored or only become obvious after the suicide occurs. Public health and mental health agencies typically only become involved when suicidal people ask for help or after an attempt occurs.
But what if it were possible to identify people at risk for suicide and provide them with help when they really need it? Though it is hardly practical (or legal) to intervene whenever someone makes a casual suicide threat, the rise of social media suggests a new way of preventing suicide deaths. Researchers have only just started exploring the potential value of Facebook, Twitter, and other social networks in identifying at-risk individuals and delivering suicide prevention information.
Facebook has already launched a new suicide prevention initiative following numerous Facebook-related deaths, many involving people posting suicide notes on the profiles. Facebook users seeing suicide messages on Facebook pages can now flag those messages for immediate intervention by suicide prevention counselors. Facebook also includes links to the National Suicide Prevention Lifeline in the U.S. to help users considering suicide.
There has been a rise in news stories related to Twitter-related suicides including a Twitter conversation in 2009 between actress Demi Moore and a fan which led to the fan's suicide being prevented. More recently, hip-hop artist Freddy E. committed suicide in 2013 after announcing his suicide intentions to fans on Twitter. Given that Twitter already has hundreds of millions of users with millions of tweets each day, it may play an important role in identifying people planning suicide.
A new research study published in the journal Crisis: The Journal of Crisis Intervention and Suicide Prevention examines how suicide risk factors can be tracked through Twitter. Jared Jashinsky and a team of fellow researchers at Brigham Young University in Provo, Utah used Twitter's application-programming interface (API) to filter Twitter conversations based on a list of suicide-related keywords. The keyword list was created based on risk factors and warning signs associated with suicide, previous suicide attempts, mental health history, and presence of guns in the home.
Search terms also included words related to bullying, substance abuse, feeling isolated, and impulsiveness. Tweets were collected and stored in an at-risk database and only those tweets that provided a clear geographic origin were included for analysis. This allowed for a state-by-state comparison of suicide-related tweets across the United States. Tweets that were either jokes, sarcastic in nature, or otherwise non-relevant were weeded out.
Sample tweets suggesting suicide risk included indications of depression ("I feel so worthless today"), substance abuse ("Dear Prozac, time for a upping in your dosage!"), prior suicide attempts ("I tried to commit suicide before…several times."), suicidal thoughts ("I have had thoughts on suicide and running away from home…and sometimes I still do."), and self-harm ("People say 'stop cutting! be happy with who you are.' its [sic] so much easier to say than do… i hate myself so much…").
Of the 1,659,274 tweets from 1,208,809 unique users throughout the worlds that were collected between May 15, 2012 and August 13, 2012, 37,717 tweets from 28,088 unique users were used for the study. As a comparison, state-by-state suicide data for 2009 was also used. This was based on the Center for Disease Control suicide statistics for that year and includes data taken from death certificates and coroner's reports. Based on correlation data comparing proportion of suicidal tweets per state with actual number of suicides, the states with the highest correlations were Alaska, New Mexico, Idaho, and South Dakota. The four lowest states were Louisiana, Maryland, Pennsylvania, and Delaware.
Overall, there does appear to be a significant correlation between rates of suicide-related tweets and actual suicide rates as measured by U.S. state. While there is no way to verify directly whether people making suicide-related tweets actually commit suicide, the social data available through. Twitter is a potentially valuable way of tracking real-time suicide risk in many users.
While teachers and counselors are trained to identify young people making statements suggesting that they are planning or in the process of attempting suicide, this is less practical for adults or adolescents who are deliberately concealing their self-harm intentions. Relying on social media data such as Facebook posts and tweets can become part of a more comprehensive strategy to "red-flag" tweets and follow them up with on-line information on available suicide resources. Users judged to be at risk for suicide could participate in Twitter conversations with trained professionals or otherwise referred to local agencies for help.
Programs using Twitter to deal with public health problems are already available. One of these, Twitcident, is a Dutch-based system to filter tweets relating to local emergencies that allows emergency service personnel to respond quickly during civil emergencies including natural disasters. Developing a similar system for suicide may be effective in both in actively preventing suicides and providing information for people who are considering it.
Jared Jashinsky and his fellow researchers point out the limitations in their study, including the risk of falsely identifying people as being at risk. As well, Twitter users are primarily younger adults (only 9 percent of users are 50 or older) so it may be less useful in identifying older people considering suicide. Still, there does appear to be a clear link between the proportion of Twitter users who may be at risk and actual suicide attempts across the United States.
As Twitter and other social media platforms continue to expand, creating new approaches using social media to reach people considering suicide can be an important part of future suicide prevention initiatives.