Data science for modeling disease interactions: a baseline algorithm

• Published in: E3S Web of Conferences 2022, Vol.351, p.1028
• Main Authors: Marzouki, Faouzi, Bouattane, Omar
• Format: Article
• Language: English
• Subjects: Aging; Algorithms; Comorbidity; Data science; Diseases; Health care; Health risks; Machine learning; Network analysis; Patients
• ISSN: 2267-1242; 2555-0403; 2267-1242
• DOI: 10.1051/e3sconf/202235101028

Multimorbidity is one of the major problems in recent health care systems, the more conditions the patients suffer from, the worst psychological pressures are put upon these patients. We formulate Multimorbidity detection as a hypergraph learning problem. Then we propose an implementation of a multimorbidity pattern detection using Multimorbidity coefficient score. This pairwise based algorithm can be considered as a baseline to which other data-driven and machine learning techniques for multimorbidity pattern detection can be evaluated. We illustrate this algorithm by building a co-occurrence model for comorbid diseases over psycho-social profiles present in a real dataset. Based on the comorbidity network of diseases, we conducted mesoscopic analysis using centrality analysis of network disease/nodes and determined potential components of the network using community detection algorithms. The patterns detected in this work by the used algorithms reveal first, that the proposed algorithm can be used as a baseline to other approaches. Second, that aging does not influence the risk of developing Multimorbidity diseases just in quantity, but also in complexity.