MorbiNet

MorbiNet constructs and analyzes multimorbidity networks to characterize comorbidity patterns and temporal disease trajectories in adult electronic health records with emphasis on type 2 diabetes mellitus (T2DM).

Key Features:

• Data Source and Scale: Uses electronic health records from Catalonia, Spain (2006–2017) comprising 3,135,948 adults, including 539,909 individuals with T2DM.
• Network Construction: Constructs multimorbidity networks using odds-ratio estimates adjusted for age and sex, with edges defined by OR > 1.2 and p < 1e-5.
• Network types and temporal trajectories: Generates undirected networks, derives directed networks, and extracts temporal disease trajectories from longitudinal associations.
• Most connected conditions: In the undirected T2DM network, complicated hypertension and atherosclerosis/peripheral vascular disease had degree 32, and cholecystitis/cholelithiasis, retinopathy, and peripheral neuritis/neuropathy had degree 31.
• Associations with high multimorbidity scores: T2DM is associated with conditions with high multimorbidity scores such as neuropathy, anemia, and digestive diseases that are linked to severe outcomes and poor prognosis.
• Directed-network associations: Directed networks show strong associations from T2DM to retinopathy (OR: 23.8), glomerulonephritis/nephrosis (OR: 3.4), peripheral neuritis/neuropathy (OR: 2.7), and pancreas cancer (OR: 2.4).
• Temporal progression findings: Temporal analyses indicate retinopathy frequently precedes complicated hypertension, cerebrovascular disease, ischemic heart disease, and organ failure.

Scientific Applications:

• Multimorbidity research: Enables characterization of comorbidity patterns and network structure in population-scale EHR data, with emphasis on T2DM.
• Risk and intervention identification: Supports identification of high-risk comorbid conditions and potential targets for intervention based on association strength and temporal ordering.
• Clinical trajectory analysis: Supports analysis of disease trajectories to inform anticipation of complications and stratification of patient management strategies.

Methodology:

Multimorbidity networks were constructed using age- and sex-adjusted odds-ratio estimates with edge inclusion threshold OR > 1.2 and p < 1e-5; directed networks and temporal trajectories were derived from longitudinal associations.