To delineate the comorbid profile of dementia through disease associations and investigate its molecular underpinnings by integrating large-scale hospital data with gene–disease associations.

Dementia is a complex clinical syndrome with high multimorbidity burden and limited disease-modifying options. Whether comorbidities cluster by chance or reflect shared biological pathways remains unclear.

We analysed >3.6 million hospital discharges of Italian adults aged ≥65 years (2015–2016), including 278,149 with dementia and 3,417,559 matched controls. Twelve chronic disease groups were derived from ICD-9-CM codes.

In step 1, all disease dyads were tested for significant co-occurrence (χ² p < 0.01, relative risk > Q3, φ > 0). In step 2, bootstrap-based comparisons identified pairs whose associations were significantly stronger in dementia than controls. In step 3, networks were normalised by internal rank and z-score to highlight links most distinctive of dementia.

Shared genes for each significant pair were retrieved from curated DisGeNET sources (OMIM, Orphanet, ClinVar, UniProt). Only overlaps with significant enrichment by Fisher’s exact test and direct connection to dementia genes were retained. Gene Ontology enrichment was then applied to identify converging biological processes.

This is the first diseasome-based analysis of dementia comorbidities integrating clinical and molecular layers. Our findings demonstrate that dementia-associated multimorbidity reflects biologically grounded rather than age-driven clustering, highlighting common vascular and hypoxia mechanisms as potential therapeutic targets.