A machine learning approach to modeling PTSD and difficulties in emotion regulation

Area of Science:

• Psychology
• Clinical Psychology
• Neuroscience

Background:

• Emotion regulation difficulties are linked to posttraumatic stress disorder (PTSD).
• Specific emotion regulation abilities most relevant to PTSD severity remain less understood.
• Understanding these links can inform targeted therapeutic interventions.

Purpose of the Study:

• To examine item-level and subscale-level models of emotion regulation difficulties in relation to PTSD severity.
• To identify specific emotion regulation deficits that are most predictive of PTSD severity.
• To utilize supervised machine learning for analyzing these relationships.

Main Methods:

• A sample of 570 U.S. adults recruited via Amazon's Mechanical Turk (MTurk).
• Self-report measures of emotion regulation difficulties (using DERS items and subscales) and PTSD severity.
• Five supervised machine learning algorithms, including ridge and elastic net regression, were employed.

Main Results:

• Emotion regulation variables explained approximately 28% (item-level) and 27% (subscale-level) of the variance in PTSD severity.
• Key item-level predictors included secondary emotional responding, emotions feeling out-of-control, difficulties modulating arousal, and low emotional granularity.
• Key subscale-level predictors were lack of access to effective strategies, lack of emotional clarity, and emotional nonacceptance.

Conclusions:

• Specific aspects of emotion regulation difficulties, particularly secondary emotional responding, emotional control, arousal modulation, emotional granularity, strategy access, clarity, and acceptance, are strongly associated with PTSD severity.
• These findings highlight the importance of targeting specific emotion regulation deficits in PTSD treatment.
• The study provides a foundation for developing more efficient and tailored interventions for individuals with PTSD.