Statements included "stop cutting myself," "I'm being bullied" and "parents fight again." To pick out only tweets that represented a true risk, the researchers had disqualifying words "I feel sad," for example, was tossed if it also contained words such as "Lakers" or "Bieber."
The authors found 37,717 troubling messages from 28,088 people. They separated the tweets by state, then compared each state's risky tweeting ratio to its suicide rate. They found a strong correlation.
Alaska, for example, has the country's highest suicide rate and also the top risky tweet ratio. Utah has the seventh-highest suicide rate and came in sixth in the study.
The concept needs refinement not every state with a high suicide rate ranked high in the Twitter analysis but BYU researchers hope it can eventually alert people far ahead of risk-assessment methods such as phone calls and surveys.
"What's so useful about this method is being able to see in real time what's going on and maybe get to people at risk a lot quicker," said co-author Carl Hanson, a BYU health scientist. "If it is indeed possible to identify in real time suicide risk, ultimately the idea is you could intervene at some point."
They're developing an app that would let people know if someone in their social circle may be suicidal.
The information is potentially useful in Utah, where state workers are also looking at social media as a prevention device, said Kim Myers, suicide-prevention coordinator at the Department of Human Services.
Still, work remains to be done to assess the best way to intervene and how to balance privacy concerns.
"What are the ethics and what are the considerations that we need to take into account when we're doing that kind of research?" Myers said. "There needs to be a strong discussion around that before we can put this kind of information to practical use."