Dynamic Bayesian network models for self-management of chronic diseases: Rheumatoid arthritis case-study

Area of Science:

• Artificial Intelligence
• Computational Biology
• Medical Informatics

Background:

• Chronic diseases require lifelong management, often with reactive interventions.
• Current management may miss sudden disease activity increases between fixed clinic visits.
• Dynamic Bayesian Networks (DBNs) offer a framework for modeling temporal event continuity.

Purpose of the Study:

• To develop a method for constructing DBN models for chronic disease self-management.
• To utilize multiple evidence sources for building these DBN models.
• To apply the approach to Rheumatoid Arthritis (RA) as a case study for treatment decision support.

Main Methods:

• Integrated rheumatology expert knowledge, clinical data, guidelines, and literature.
• Identified variables, states, dependencies, and parameters for the DBN.
• Employed data frequency manipulation and dummy patient scenarios for initial model inference due to data limitations.

Main Results:

• The developed DBN for RA exhibited a clinically meaningful structure.
• Initial evaluations showed promising results for informing treatment decisions.
• The approach successfully integrated diverse evidence sources to build the DBN.

Conclusions:

• The proposed method effectively builds DBN models for chronic disease self-management.
• The RA DBN model shows potential as a decision-support tool for patients and clinicians.
• Further evaluation and calibration are needed for clinical implementation.