Modeling comorbidity of chronic diseases using coupled hidden Markov model with bivariate discrete copula

Area of Science:

• Biostatistics
• Epidemiology
• Computational Biology

Background:

• Chronic diseases often interact and share risk factors, a phenomenon known as comorbidity.
• Comorbidity is crucial for accurate actuarial valuations and understanding disease progression.
• Existing models may not fully capture the complex dynamics of interacting chronic conditions.

Purpose of the Study:

• To develop a novel statistical model for analyzing parallel, interacting chronic disease processes.
• To introduce a coupled hidden Markov model (CHMM) integrated with bivariate discrete copula functions.
• To provide a method for estimating model parameters and addressing computational challenges.

Main Methods:

• Utilized a combination of hidden Markov theory and copula functions to model disease interactions.
• Developed a CHMM with a bivariate discrete copula within the hidden process.
• Employed a variational expectation-maximization (VEM) algorithm for parameter estimation.
• Addressed numerical underflow issues in forward-backward probability computations.

Main Results:

• Simulation studies demonstrated satisfactory performance of the proposed CHMM across different association levels.
• The model successfully defined the dependency structure and dynamics of unobserved disease data.
• Application to hospital appointment data showed the model's utility in analyzing comorbidity.

Conclusions:

• The proposed CHMM offers a robust framework for investigating disease comorbidity.
• The model is particularly valuable when analyzing population dynamics over time without access to clinical data.
• This approach enhances our understanding of complex disease interactions and their impact.