Structural FHMM Reveals Interpretable T2DM Disease Trajectories.

Understanding the complex progression of chronic diseases like Type 2 Diabetes Mellitus (T2DM) is crucial for personalized patient care. Traditional models often struggle to provide interpretable insights into the multifaceted nature of disease trajectories. This new research introduces a structural variant of the Factorial Hidden Markov Model (FHMM) designed specifically for analyzing T2DM progression. The proposed model represents a patient's underlying health state not as a single entity, but as a combination of multiple independent components that evolve simultaneously. These components are associated with various comorbidities and lab results, allowing for a more granular and clinically meaningful understanding of the disease. This structured latent representation facilitates the identification of distinct patient states and the clustering of common disease trajectories. Evaluated using anonymized electronic health records from a large database, the model successfully identified multiple clinically coherent latent components corresponding to known patterns of diabetes-related complications. It revealed heterogeneous progression pathways, including distinct microvascular-dominant and multi-organ trajectories, which were linked to elevated comorbidity burden and mortality. These findings demonstrate the framework's ability to extract meaningful longitudinal structures from EHR data and provide interpretable insights into T2DM evolution.