UNRAVELLING THE COMPLEXITY OF COMORBIDITIES AND POLYPHARMACY IN EPILEPTIC PATIENTS FROM PATIENT-GENERATED DATA: A MACHINE LEARNING DRIVEN STUDY

Author: Elavarasi Jeyakumar
Degree: MSc Health Data Science and Statistics
Institution: University of Plymouth
Year: 2025

Overview

Large-scale analysis of 20,146 participants from the US National Health and Nutrition Examination Survey (NHANES) 2013-2016 to uncover comorbidity and polypharmacy patterns in 828 epilepsy patients.

Key Findings

• 100% of epilepsy patients had at least one additional chronic condition
• 75.5% classified as high-risk using composite risk stratification model
• Strong correlation (r=0.476, p<0.001) between comorbidity burden and medication count
• Three distinct patient phenotypes identified via K-means clustering

Methods & Tools

• Python (pandas, scikit-learn, matplotlib)
• K-means clustering with PCA
• Statistical testing: Mann-Whitney U, Fisher's exact test, Spearman correlation
• Risk stratification modelling
• NHANES 2013-2016 datasets

Files

• Untitled5.ipynb — full analysis code with outputs
• dissertation pdf — full research report

Contact

www.linkedin.com/in/elavarasi-jeyakumar-4526403b9