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

Area of Science:

• Biomedical Informatics
• Network Science
• Computational Biology

Background:

• Current multimorbidity research often relies on count scores, disease clustering, and pairwise associations.
• These methods fail to reveal underlying biological mechanisms and overlook higher-order interactions like effect modification.

Purpose of the Study:

• To introduce novel multimorbidity metrics by integrating signs, symptoms, and diseases into comprehensive networks.
• To address the limitations of pairwise associations by incorporating higher-order interactions, leading to the development of hypergraphs.

Main Methods:

• A synthetic Bayesian Network model incorporating pairwise and higher-order interactions was developed.
• Network interventions were simulated at individual and population levels.
• Outcomes were compared against predictions derived solely from pairwise associations.

Main Results:

• Pairwise association analysis can lead to missed intervention opportunities or unexpected side effects.
• Hypergraphs reveal crucial links overlooked in traditional pairwise networks.
• A more complete overview of sign and disease associations is achieved using hypergraphs.

Conclusions:

• Higher-order interactions are essential for a comprehensive understanding of multimorbidity.
• Hypergraphs offer a superior framework for analyzing complex disease relationships compared to pairwise networks.
• This approach enhances the accuracy of predicting intervention outcomes in multimorbidity research.