🤖 AI Summary
Existing approaches struggle to effectively model the complex interactions among disease progression, multimorbidity networks, and high-dimensional time-varying risk factors. This work proposes a structured Bayesian continuous-time Bayesian network to learn directed disease dependencies from longitudinal electronic health records, wherein transition intensities depend on existing conditions, pairwise interactions, and exogenous covariates. To mitigate parameter explosion in higher-order terms while preserving interpretability of main effects, the model incorporates order-dependent structured shrinkage priors—such as spike-and-slab and Bayesian LASSO—that selectively suppress redundant interactions. Simulation studies demonstrate that the spike-and-slab prior achieves superior performance in variable selection, network recovery, and false discovery control. Applied to UK Biobank data, the method successfully identifies a diabetes-centered metabolic module and a respiratory–atopic inflammatory module.
📝 Abstract
Multiple long-term conditions (MLTCs) arise through complex, time-dependent interactions among diseases, yet existing methods often struggle to jointly model disease progression, multimorbidity networks, and high-dimensional risk factors. We propose a structured Bayesian continuous-time Bayesian network (CTBN) framework for learning directed disease-dependency networks from longitudinal electronic health records. The model allows disease transition intensities to depend on existing conditions, pairwise disease interactions, and exogenous covariates. To control the combinatorial growth of interaction parameters, we introduce order-dependent shrinkage priors that increasingly penalise higher-order effects while preserving clinically interpretable main effects. We compare four sparsity-inducing priors, spike-and-slab, structured normal, Bayesian LASSO, and regularised horseshoe through extensive simulation studies. Across multiple data-generating scenarios, the spike-and-slab prior achieved the best network recovery, variable-selection accuracy, and false-discovery control, while continuous shrinkage priors were less effective for hard variable selection. The proposed framework was applied to UK Biobank primary care records, focusing on data from 33,558 participants who were free of the ten selected most prevalent conditions at age 40 and who subsequently developed at least one of these conditions during the follow-up period. The selected spike-and-slab model identified two dominant disease modules: a cardiometabolic cluster centred on diabetes and an inflammatory cluster linking respiratory and atopic conditions.