Disease Cluster Analysis in Electronic Health Records (MIMIC-IV)
Cohort of 179,524 adult patients (ages 18–91), 11,733 mortality cases, 846 distinct diseases. Apriori association rule mining (minimum support 0.01) found 628 association rules with confidence above 50%. k-means clustering after t-SNE, k = 4 by silhouette analysis (score 0.44). Published in IISE Annual Conference & Expo proceedings, 2025.
Timeline
2025
Impact Focus
628 association rules with confidence above 50%; published in IISE Annual Conference & Expo proceedings, 2025.
Core Tools
EHR, MIMIC-IV, Python, clustering
Problem
Comorbidity and mortality patterns in electronic health records.
What I did
Apriori association rule mining (minimum support 0.01), t-SNE, and k-means clustering; k = 4 by silhouette analysis (score 0.44).
Result
628 association rules with confidence above 50%; published in IISE Annual Conference & Expo proceedings, 2025.
What changed
- - 179,524 adult patients (ages 18–91)
- - 11,733 mortality cases and 846 distinct diseases
Deliverables
- - 628 association rules with confidence above 50%; published in IISE Annual Conference & Expo proceedings, 2025.