SSRN Electronic Journal · Jan 2026

AI-Powered Phenotypic Clustering Reveals a Dementia-Lipids-Vascular Network and a Dual Vitamin-Cardiovascular Axis

By Stoenescu S, Azarmaneche M

Background: Network Medicine posits that diseases form interconnected topological clusters, but verifying these clusters in clinical data requires filtering the noise of administrative coding. MEAT criteria (Monitor, Evaluate, Assess, Treat) provide a rigorous standard for distinguishing active disease management from background documentation. Objective: To discover cross-system syndromic clusters in a large geriatric population using Association Rule Mining on MEAT-verified diagnoses at the ICD-10 category level, with particular focus on neurodegenerative-metabolic and vitamin-cardiovascular axes. Methods: We analyzed 104,724 patients (mean age 71.6 ± 12.1 years, 54.6% female) drawn from a clinically validated dataset previously validated against CDC national benchmarks [1]. ICD-10 codes were grouped at the 3-character category level. Encounter/status (Z), obstetric (O), and external cause (V-Y) codes were excluded. The Apriori algorithm identified frequent 2-and 3-itemsets; Lift, Confidence, and chi-square metrics quantified association strength. Results: From 170 ICD-10 categories meeting 0.5% prevalence, we identified 1,024 frequent pairs and 1,892 frequent triplets. Two novel cross-system networks emerged: (1) a Dementia-Lipids-Vascular Network linking Alzheimer's disease (G30), unspecified dementia (F03), hyperlipidemia (E78), and hypertension (I10), with triplet Lift values of 15.9-18.7 (χ 2 > 9,000, p < .001); and (2) a Dual Vitamin-Cardiovascular Axis linking B-vitamin deficiency (E53) and Vitamin D deficiency (E55) jointly to coronary artery disease (I25, Lift 12.6), hypertension (I10, Lift 7.7), and chronic kidney disease (N18, Lift 10.5). The previously reported Metabolic-Bone-Vascular axis was confirmed (Lift 2.1-2.4), and a Metabolic Spine phenotype (E78+M47+M51, Lift 9.6) extended the metabolic-musculoskeletal hypothesis beyond the knee. Conclusions: MEAT-verified network analysis uncovers high-fidelity biological clusters that challenge organ-system segregation. The Dementia-Lipids-Vascular Network supports vascular contributions to neurodegeneration. The Dual Vitamin-Cardiovascular Axis identifies a two-pathway nutritional risk signature for atherosclerosis.

doi.org/10.2139/ssrn.6639398Open accessPopulations

AgingResearch.News