Inferring Trajectories of Psychotic Disorders Using Dynamic Causal Modeling.

Area of Science:

• Psychiatry
• Computational Neuroscience
• Data Science

Background:

• Current psychiatric nosologies have limited ability to capture the complexity of illness courses.
• Existing models often overlook detailed temporal dynamics of symptoms.
• Understanding illness trajectories is crucial for accurate diagnosis and treatment.

Purpose of the Study:

• To introduce and validate a Dynamic Causal Model (DCM) for characterizing detailed psychiatric illness course patterns.
• To assess the DCM's ability to model symptom trajectories using dense time-series data.
• To evaluate the DCM's accuracy in estimating latent course patterns and distinguishing between them.

Main Methods:

• A three-level DCM was constructed to model latent dynamics underlying depression, mania, and psychosis symptoms.
• The model was applied to prospective symptom scores from nine patients over four years.
• Model validation involved simulations and group-level analyses using Parametric Empirical Bayes (PEB) and cross-validation.

Main Results:

• The DCM accurately captured individual symptom trajectories for all nine patients.
• Simulations demonstrated accurate parameter estimation (correlations >= 0.76).
• The DCM successfully distinguished between different latent course patterns, with PEB correctly assigning 8 out of 9 simulated patients.

Conclusions:

• Dynamic Causal Modeling (DCM) offers a powerful method for analyzing complex symptom trajectories in psychiatric disorders.
• This approach can explicate temporal patterns defining nosologic entities.
• The findings suggest potential for DCM in facilitating personalized psychiatric treatment.