Machine Learning To Provide a Gold Standard in PTSD Diagnosis

Machine-Learning Diagnostic Tool

Researchers from the VA Boston Healthcare System and the Boston University School of Public Health (BUSPH) were able to use machine learning to diagnose post-traumatic stress disorder (PTSD) more efficient.

The diagnosis of PTSD consumes a lot of time, which usually takes 30 minutes or more, a process that can be considered too long for a routine clinical visit.

The researchers developed a machine learning tool to explore the streamlining of the gold standard tool in diagnosing PTSD.

Eight million adults in the US are affected with PTSD, including the hundreds of thousands of veterans from the Iraq and Afghanistan conflicts. These PTSD symptoms are also on the rise among the general population amidst the COVID crisis as everyone's mental health is affected by this pandemic.

READ ABOUT: Artificial Intelligence Can Detect Brain Blockages and Reduce Time to Treatment

While studying the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders, fifth edition (SCID-5), the researchers found that they can remove six of the 20 questions while maintaining at least 90 percent accuracy for the general veteran population. This was published in the journal Assessment.

Data Assessment

The team used the data from the manual to assess the 1,265 veterans who served in Iraq and Afghanistan. They then built a machine learning model called "random forests," as it is made up of "forests" of decision trees to identify how different terms have weaker associations in PTSD diagnosis and can be cut instead.

Tammy Jiang, a lead author of the study and doctoral candidate in epidemiology at BUSPH, said that it was through this data that they found some of the PTSD items can be removed due to its redundancy. As it does not make any substantial contributions or specific enough to predict PTSD relative to the other PTSD question accurately, they can be removable.  

The group of researchers found that for these veterans, the following items can be removed: dissociative reactions, reckless or self-destructive behavior, irritable behavior and anger, hypervigilance, persistent inability to experience positive emotions, and exaggerated startle response.

The veteran sample consists of half male and half female and found that the essential item for a diagnosis was detachment or estrangement from others. This item applies to the whole sample for both male and female veterans separately. 

However, for male and female veterans, different items can be cut.

For the male veterans, these four items can be removed: inability to recall important aspects of a traumatic event, dissociative reactions; reckless or self-destructive behavior; and hypervigilance.

For the female veterans, six items were identified: reckless or self-destructive behavior; dissociative reactions; persistent inability to experience positive emotions; irritable behavior and angry outbursts; exaggerated startle response; and hypervigilance.

Dr. Brian Marx, senior author of the study and works as a staff psychologist at VA Boston Healthcare System, said that the research demonstrates how the diagnosis of PTSD differs among men and women.  

Machine-Learning Not a Replacement for Healthcare Providers

Even with all these promising results, the researchers emphasized that machine learning algorithm is not meant to replace human mental health providers, but to act as a companion tool to provide a more efficient PTSD diagnosis, without compromising the quality of care provided.

The study's findings have paved a path where they can skip to a diagnostic interview, which is abbreviated but still the "gold standard," and can accurately identify people who have PTSD so they can get treatment as early as possible.

SEE HERE: Antidepressant Drug Could Potentially Treat Lazy Eye 

Check out more news and information on PTSD on MD News Daily.

© MD News Daily.