New research suggests that speech patterns may reveal a person’s risk for psychosis-related disorders. The discovery could lead to earlier diagnosis.
Identifying which at-risk young people will develop psychotic disorders can be a frustrating guessing game for mental health experts.
But new technology that can analyze speech patterns is raising hopes that, in the future, identifying those at risk for psychosis will be as easy as having a conversation.
A small study out this week found that a computer algorithm could identify who would develop psychosis with an accuracy of up to 83 percent.
Psychosis is a frightening condition that’s “characterized as disruptions to a person’s thoughts and perceptions that make it difficult for them to recognize what is real and what isn’t,” according to the National Alliance on Mental Illness.
Psychosis can be caused by a host of mental health conditions such as schizophrenia, which is a psychotic disorder, as well as depression and bipolar disorder.
While there are known risk factors, such as having a family member with a psychotic disorder, mental health experts haven’t been able to determine who among those at risk will actually develop psychosis.
In recent years, researchers have turned to computer algorithms to help them parse the language of at-risk individuals to see if there are clues in their speech.
This week, researchers reported in a small study that speech patterns may help reveal who is likely to develop psychosis.
Researchers from Icahn School of Medicine at Mount Sinai, the New York State Psychiatric Institute, the University of California Los Angeles (UCLA), and other institutions used a computer algorithm to examine the speech patterns of 93 at-risk young people in New York and California.
Their results were published this week in World Psychiatry.
The computer analyzed transcripts of interviews with the subjects that had been conducted years earlier.
Words were coded so that the algorithm could determine which words were out of place. As a result, the program could figure out when a person likely went off topic during the interview.
Researchers said the algorithm could identify which patients went on to develop psychosis with 83 percent accuracy. The team then used the program on a second group of study patients and found it had a 79 percent accuracy rate.
The program could also differentiate between healthy people and those with recent psychosis onset with 72 percent accuracy.
Dr. Cheryl Corcoran, associate professor of psychiatry at the Icahn School of Medicine at Mount Sinai and co-author of the study, said that if people were prone to losing the thread of the conversation they appeared to be more at risk to develop psychosis.
“The ones who went on to develop schizophrenia… they were tangential they had this language impairment,” she said.
Corcoran said the computer was able to identify these tangential breaks more deftly than most researchers.
“They detect a pattern when the topic changes,” she said. “The computer can do a much more nuanced analysis of language.”
Corcoran said it’s important to develop better techniques to identify people who will develop psychosis.
Today, mental health experts can determine who is likely to develop psychosis by looking at their current symptoms, but many of these at-risk people will not develop a full-blown psychotic disorder.
Corcoran said of those people with psychosis risk factors, “about 20 percent develop a psychotic disorder.”
Corcoran hopes this kind of research will eventually be turned into a screening tool. “My hope is that we can use this to screen individuals and if it seems like they are at risk for psychosis we can evaluate them and offer them treatment,” Corcoran said.
Other mental health experts said that the study builds on new research where speech and language are examined for signs of who is at risk.
Dr. Michael Birnbaum, of the early treatment program at Zucker Hillside Hospital in New York, said if these results are confirmed in further studies it would “be a game change.”
“I am a big fan,” Birnbaum told Healthline. “I think this could absolutely help, and essentially the study is suggesting that there are subtle language patterns that could be detected through machine learning algorithms.”
Dr. Ramani Durvasula, a professor of psychology at California State University Los Angeles, said the study was “very, very interesting.”
Durvasula said if these people can be identified early, they could be targeted to receive more specified education about the condition and undergo more frequent psychiatric check-ins to manage issues like stress.
“This has always been a dream of all mental health practitioners: prevention,” she said. “Once a problem is on board, now we’re trying to get ahead of the fire.”
While there is no “silver bullet” cure for psychotic disorders such as schizophrenia, Durvasula and Corcoran said there are many ways mental health experts can assist those at risk.
Corcoran said experts can use cognitive behavioral therapy and close monitoring to help.
“We encourage people to spend less time alone because individuals who are at risk for psychosis tend to isolate themselves,” Corcoran said. “It’s better to be with other people.”
While Durvasula found the research promising, she also pointed out that it needs to be proven to work in further studies and across different languages and cultures before being used widely by mental health professionals.
“The only potential dark side is until they’re really confident of the reliability and validity of this system, we always have to be careful because issues like psychosis are stigmatizing,” said Durvasula. “If we start slapping potential labels on people, then the data really have to hold.”