CINCINNATI: A research team at an Ohio children’s hospital has been analyzing a collection of more than 1,300 suicide notes to try to help save lives.
John Pestian, the director of computational medicine at Cincinnati Children’s Hospital Medical Center, and his team are using advanced computer technology to analyze the language in the notes written by people of varying ages from all over North America.
The team hopes to gain a better understanding of the writers and use that information to create a tool that can help mental health workers assess the likelihood a person will attempt suicide, The Cincinnati Enquirer (http://cin.ci/ZUmrHf) reported.
Only a few other suicide researchers are merging psychology and computational analysis, and Pestian is “really doing some groundbreaking work,” said Michelle Linn-Gust, president of the American Association of Suicidology.
The efforts are especially important because research to prevent suicide has reached a plateau, and the problem “is not going away,” according to Linn-Gust.
Someone dies by suicide every 14 minutes in the United States, according to the newspaper.
About 40 suicidal young people come to the emergency department at Cincinnati Children’s each week, Pestian said.
Assessing a person’s risk for suicide often falls to social workers, nurses, psychologists or doctors — professionals whose training and life experiences can vary greatly.
“People hear things differently,” and that can result in differences in determining whether someone is suicidal, said Pestian.
His team is trying to develop more support for people making those decisions.
Pestian, who works in the field of neuropsychiatric computational linguistics, said he teaches “computers how to listen, and report back what they think they’ve heard from people.”
The first step in the current research was to collect the notes written by people who died by suicide.
“When people hear you’re doing this work, they step up,” he said.
Surviving family members from across North America sent notes that were written between 1950 and the present by people of all ages and those notes can be very depressing, Pestian said.
“I’ve cried more times than I can count over some of the things I’ve read,” he said. “But in the end, you’re doing whatever you can to help save lives, to help the human condition.”
The notes were scanned, transcribed and reviewed for accuracy and 165 volunteers — people who had lost loved ones to suicide — were recruited to read them and select words, phrases or sentences that represent emotions including anger, fear, guilt, hopelessness and love — among others.
Results were put into a database against which computers could compare the speech of people at risk of suicide, and a set of instructions was created to teach a computer how to find patterns and make predictions from the data.
The set of instructions — an algorithm — was tested in a clinical trial by asking a series of questions to 30 young people with suicidal tendencies and 30 in a control group.
“We wanted to know if the computer could tell, by listening to recordings of what they said, which ones are suicidal, and which ones aren’t,” Pestian said.
He said the computer was 93 percent accurate in identifying those with suicidal tendencies, while humans were right slightly more than 50 percent of the time with the same groups.
The team plans a larger experiment at more sites that could lead to a product for use in a clinical setting, possibly within a couple of years, Pestian said.
But he noted that such a product wouldn’t replace the importance of people in making patient assessments.
Ultimately, he said, “the clinician, the person at the bedside, makes the decisions.”