Understanding multimorbidity requires sign-disease networks and higher-order interactions, a perspective

Background: Count scores, disease clustering, and pairwise associations between diseases remain ubiquitous in multimorbidity research despite two major shortcomings: they yield no insight into plausible mechanisms underlying multimorbidity, ignore higher-order interactions such as effect modification. Objectives: We argue that components are currently missing but vital to develop novel metrics. Firstly, networks should be constructed which consists simultaneously of signs, symptoms, diseases, since only then could shared biological diseases. Secondly, learning is insufficient fully characterize the correlations a system. That is, synergistic (e.g., cooperative or antagonistic) effects widespread complex systems, where more elements combined give larger smaller than sum their individual effects. It can even occur pairs symptoms have whatsoever, combination significant association. Therefore, included used study resulting so-called hypergraphs. Methods: illustrate our argument using synthetic Bayesian Network model signs composed interactions. simulate network interventions on both population levels compare ground-truth outcomes with predictions from associations. Conclusion: find that, when judged purely associations, unexpected “side-effects” most opportune intervention missed. The hypergraph uncovers links missed networks, giving complete overview sign