Working with 9/11 responders as retrospective subjects, researchers in mental health and computer science have shown AI language tools can help predict the course of PTSD.
The tools were trained on and applied to transcripts of clinical interviews. The algorithms proved helpful making both long-term and point-in-time predictions.
Based on this two-tiered success, the researchers suggest AI may be helpful not only for refining psychiatric therapy but also for screening in primary care.
The project was conducted at Stony Brook University in New York and is described in a study published June 22 in Psychological Medicine.
PhD candidate Youngseo Son and colleagues focused on the oral histories of 113 individuals who responded to the 9/11 attacks on the World Trade Center and had at least one PTSD diagnosis within two years of their initial interview.
The team used an established PTSD checklist to measure symptom severity for up to seven years after the initial interview.
They computed AI-based indicators for depression, anxiety and other mental-health problems, and they incorporated “dictionary-based measures of linguistic and interpersonal style,” according to the study report.
They found that, cross-sectionally, greater depressive language and first-person singular usage correlated with increased symptom severity.
Longitudinally, meanwhile, anxious language predicted future worsening in PTSD checklist scores, while first-person plural usage and use of longer words predicted improvement.
“This is the first study to demonstrate the value of AI in understanding PTSD in a vulnerable population,” write Son, senior author H. Andrew Schwartz, PhD, and co-authors. “Future studies should extend this application to other trauma exposures and to other demographic groups, especially under-represented minorities.”
In addition, they comment, the patterns of language-based assessments consistent with previous findings in other settings and their strong statistical correlations …
provided unique insights and explanations beyond commonly known confounds or risk factors such as age, gender, occupation, marital status or even questionnaire-based depression measures, suggesting support for clinicians toward more precise decisions.
More generally, language-based assessments that capture individual digital phenotypes and distinctive linguistic markers from transcripts of interviews are very useful for investigating underlying causes of PTSD and may play a critical role as a supplement for enhancing personalized preventive care and more effective treatments for PTSD; they may even enable real-time screening or preventive measures with reduced costs and less therapist time for helping a large number of people exposed to large-scale traumatic events similar to a previous online PTSD treatment.”
The study is available in full for free.
Dave P. has worked in journalism, marketing and public relations for more than 30 years, frequently concentrating on hospitals, healthcare technology and Catholic communications. He has also specialized in fundraising communications, ghostwriting for CEOs of local, national and global charities, nonprofits and foundations.