Classification and discrimination of emotion dysregulation disorders using machine learning

Area of Science:

• Psychiatry
• Neuroscience
• Machine Learning

Background:

• Attention-deficit/hyperactivity disorder (ADHD), bipolar disorder (BD), and borderline personality disorder (BPD) are severe psychiatric conditions.
• These disorders frequently exhibit overlapping symptoms of emotion dysregulation.
• The potential for a continuum of emotion dysregulation across these disorders remains unclear.

Purpose of the Study:

• To investigate whether ADHD, BD, and BPD exist on a continuum of emotion dysregulation.
• To utilize machine learning (ML) to predict and discriminate between these disorders using psychological and environmental features.
• To examine how offspring of individuals with these disorders relate to their parent disorders and healthy controls.

Main Methods:

• Machine learning algorithms were applied to data from 232 adults, including 92 patients with ADHD, BD, or BPD, 67 genetically unrelated offspring, and 73 healthy controls.
• Features analyzed included clinical dimensions of emotion dysregulation (e.g., impulsivity, mania, rumination), childhood trauma, and parental bonding.
• Classification and prediction models were developed to differentiate between the diagnostic groups.

Main Results:

• Borderline personality disorder (BPD) demonstrated strong discrimination from healthy controls and ADHD, and moderate discrimination from bipolar disorder (BD).
• Attention-deficit/hyperactivity disorder (ADHD) showed strong discrimination from healthy controls and BPD, and moderate discrimination from BD.
• Bipolar disorder (BD) exhibited moderate discrimination from ADHD and BPD, and weak discrimination from healthy controls.
• Offspring groups more closely resembled healthy controls than their respective parent disorders.
• Machine learning reliably classified ADHD, BD, and BPD as independent constructs based on emotion regulation and environmental factors.

Conclusions:

• ADHD, BD, and BPD represent distinct psychiatric constructs, not a continuum of emotion dysregulation.
• Machine learning, utilizing emotion regulation traits and environmental factors, offers a reliable method for classifying these disorders.
• The findings suggest potential for cost-effective ML-based classification and prediction of these emotion dysregulation disorders.