Spreading of diseases through comorbidity networks across life and gender

TL;DR

This study constructs a comprehensive comorbidity network from Austrian medical claims data, revealing age- and gender-dependent structural changes and disease spreading patterns, with implications for predicting disease progression and prevention.

Contribution

It introduces a multiplex disease network based on population data, showing how disease structures evolve with age and gender, and models disease spreading dynamics.

Findings

Disease networks are highly interconnected in children.

Distinct disease clusters emerge during adolescence and adulthood.

In the elderly, disease hubs like hypertension dominate the network.

Abstract

The state of health of patients is typically not characterized by a single disease alone but by multiple (comorbid) medical conditions. These comorbidities may depend strongly on age and gender. We propose a specific phenomenological comorbidity network of human diseases that is based on medical claims data of the entire population of Austria. The network is constructed from a two-layer multiplex network, where in one layer the links represent the conditional probability for a comorbidity, and in the other the links contain the respective statistical significance. We show that the network undergoes dramatic structural changes across the lifetime of patients.Disease networks for children consist of a single, strongly inter-connected cluster. During adolescence and adulthood further disease clusters emerge that are related to specific classes of diseases, such as circulatory, mental, or genitourinary disorders.For people above 65 these clusters start to merge and highly connected hubs dominate the network. These hubs are related to hypertension, chronic ischemic heart diseases, and chronic obstructive pulmonary diseases. We introduce a simple diffusion model to understand the spreading of diseases on the disease network at the population level. For the first time we are able to show that patients predominantly develop diseases which are in close network-proximity to disorders that they already suffer. The model explains more than 85 % of the variance of all disease incidents in the population. The presented methodology could be of importance for anticipating age-dependent disease-profiles for entire populations, and for validation and of prevention schemes.