Depression is Associated with Treatment Response Trajectories in Adults with Prolonged Grief Disorder: A Machine...

Area of Science:

• Psychiatry and Behavioral Health
• Computational Psychiatry
• Clinical Psychology

Background:

• Evidence-based treatments for Prolonged Grief Disorder (PGD) are available, but patient characteristics influencing treatment response remain unclear.
• Optimizing PGD treatment requires identifying pretreatment factors associated with differential improvement trajectories.
• Machine learning offers a novel approach to analyze complex patient data and predict treatment outcomes.

Purpose of the Study:

• To identify pretreatment clinical factors that predict distinct treatment response trajectories in patients with Prolonged Grief Disorder (PGD).
• To leverage machine learning to analyze data from a randomized clinical trial of PGD treatments.
• To inform the development of individualized treatment strategies for PGD.

Main Methods:

• Unsupervised and supervised machine learning, including latent growth mixture modeling and logistic regression with elastic net regularization, were employed.
• Data from 333 PGD patients (aged 18-95) randomized to citalopram or placebo with either grief-informed clinical management or prolonged grief disorder therapy (PGDT) were analyzed.
• Symptom trajectories were assessed using the Inventory for Complicated Grief over 20 weeks.

Main Results:

• Three distinct response trajectories were identified: lesser severity responders (60%), greater severity responders (18.02%), and non-responders (21.92%).
• The machine learning model showed acceptable discrimination between responders and non-responders (AUC = .702, accuracy = .684).
• Higher baseline depression severity, greater grief-related functional impairment, and lack of PGDT were associated with a lower likelihood of treatment response.

Conclusions:

• Machine learning can effectively identify distinct PGD treatment response patterns.
• Early identification of higher depression severity, functional impairment, and absence of PGDT is crucial for optimizing PGD treatment.
• These findings support the need for personalized PGD treatment strategies based on pretreatment patient characteristics.