Comorbid bipolar disorder and borderline personality disorder: Diagnosis using machine learning

Area of Science:

• Psychiatry
• Machine Learning
• Clinical Psychology

Background:

• Comorbid bipolar disorder (BP) and borderline personality disorder (BPD) poses significant diagnostic challenges.
• Accurate differentiation is crucial for effective treatment planning and patient outcomes.

Purpose of the Study:

• To employ a machine learning (ML) approach for distinguishing comorbid BP/BPD from BP and BPD individually.
• To identify key clinical variables that differentiate these conditions.

Main Methods:

• Utilized machine learning algorithms on participant data with DSM diagnoses.
• Compared 82 BP, 52 BPD, and 53 comorbid BP/BPD participants.
• Analyzed self-report measures of personality, emotion regulation, and childhood parental experiences.

Main Results:

• ML achieved 79.6% accuracy in classifying BP/BPD vs. BP and 61.7% vs. BPD.
• Stress-related ideation and borderline personality traits distinguished BP/BPD from BP.
• Emotion regulation deficits were key in differentiating BP/BPD from BPD.
• Impulsivity and anger were significant across both comparisons.

Conclusions:

• Identified distinct clinical variables for comorbid BP/BPD, improving diagnostic clarity.
• The ML model shows potential for enhancing clinical decision-making in complex cases.
• Further research with larger datasets is recommended to refine ML diagnostic accuracy.