Observational study reveals multi-resolution disease clusters in primary care patients, highlighting structured co-occurrence patterns across multimorbidity.
Key Points
Markov multiscale community detection outperforms k-means across 253 known disease pairs, uncovering almost-hierarchical disease clusters.
Analysis of primary care records extracts representations of 212 conditions across ten million patients using skip-gram and co-occurrence models.
Supports future discovery of novel disease associations from electronic health records, facilitating targeted clinical management of multimorbidity.