A Functional Model for Structure Learning and Parameter Estimation in Continuous Time Bayesian Network: An Application in Identifying Patterns of Multiple Chronic Conditions

Bayesian networks are powerful statistical models to study the probabilistic\nrelationships among set random variables with major applications in disease\nmodeling and prediction. Here, we propose a continuous time Bayesian network\nwith conditional dependencies, represented as Poisson regression, to model the\nimpact of exogenous variables on the conditional dependencies of the network.\nWe also propose an adaptive regularization method with an intuitive early\nstopping feature based on density based clustering for efficient learning of\nthe structure and parameters of the proposed network. Using a dataset of\npatients with multiple chronic conditions extracted from electronic health\nrecords of the Department of Veterans Affairs we compare the performance of the\nproposed approach with some of the existing methods in the literature for both\nshort-term (one-year ahead) and long-term (multi-year ahead) predictions. The\nproposed approach provides a sparse intuitive representation of the complex\nfunctional relationships between multiple chronic conditions. It also provides\nthe capability of analyzing multiple disease trajectories over time given any\ncombination of prior conditions.\n