Representing multimorbid disease progressions using directed hypergraphs

Abstract

ObjectiveTo introduce directed hypergraphs as a novel tool for assessing the temporal relationships between coincident diseases, addressing the need for a more accurate representation of multimorbidity and leveraging the growing availability of electronic healthcare databases and improved computational resources.MethodsDirected hypergraphs offer a high-order analytical framework that goes beyond the limitations of directed graphs in representing complex relationships such as multimorbidity. We apply this approach to multimorbid disease progressions observed from two multimorbidity sub-cohorts of the SAIL Databank, after having been filtered according to the Charlson and Elixhauser comorbidity indices, respectively. After constructing a novel weighting scheme based on disease prevalence, we demonstrate the power of these higher-order models through the use of PageRank centrality to detect and classify the temporal nature of conditions within the two comorbidity indices.ResultsIn the Charlson population, we found that chronic pulmonary disease (CPD), cancer and diabetes were conditions observed early in a patient’s disease progression (predecessors), with stroke and dementia appearing later on (successors) and myocardial infarction acting as a transitive condition to renal failure and congestive heart failure. In Elixhauser, we found renal failure, neurological disorders and arrhythmia were classed as successors and hypertension, depression, CPD and cancer as predecessors, with diabetes becoming a transitive condition in the presence of obesity and alcohol abuse. The dynamics of these and other conditions changed across age and sex but not across deprivation. Unlike the directed graph, the directed hypergraph could model higher-order disease relationships, which translated into stronger classifications between successor and predecessor conditions, alongside the removal of spurious results.ConclusionThis study underscores the utility of directed hypergraphs as a powerful approach to investigate and assess temporal relationships among coincident diseases. By overcoming the limitations of traditional pairwise models, directed hypergraphs provide a more accurate representation of multimorbidity, offering insights that can significantly contribute to healthcare decision-making, resource allocation, and patient management. Further research holds promise for advancing our understanding of critical issues surrounding multimorbidity and its implications for healthcare systems.